Meningeal lymphatic dysfunction mediates postoperative sleep deprivation-induced cognitive decline via impaired neuroinflammatory clearance

Abstract

Postoperative sleep deprivation (PSD) exacerbates perioperative neurocognitive disorders (PND) through unresolved neuroinflammation, yet the underlying mechanisms remain elusive. Here, we aim to evaluate contribution of disregulation of the meningeal lymphatic vessels (MLVs) to PSD-induced cognitive decline. In a mosue model combining tibial fracture surgery with sleep deprivation, PSD induced spatial and contextual memory deficits, concurrent with hippocampal neuroinflammation and microglial activation. Crucially, PSD induced significant MLVs structural regression and functional impairment, which further limited drainage of inflammatory mediators (IL-1β, IL-6, and TNF-α) to deep cervical lymph nodes. Lymphatic ligation that suppressed cytokine trafficking to peripheral lymphatics amplifies hippocampal inflammation. Overexpression of vascular endothelial growth factor C (VEGF-C) by intracisternal delivery of AAV-VEGF-C to enhance MLVs function restored lymphatic drainage capacity, attenuated hippocampal neuroinflammation, and rescued PSD-induced cognitive decline. Our findings exhibit a MLVs-related pathological mechanism in PSD-related cognition impairment, showing that sleep loss disrupts meningeal lymphatic integrity, which further impairs inflammatory clearance and exaggrates neuroinflammation and neuronal dysfunction. Therapeutic targeting of restore MLVs’ function may thus offer a novel strategy to mitigate PND.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12974-026-03784-5.

Introduction

Perioperative Neurocognitive Disorders (PND), encompassing cognitive impairments from delirium to delayed neurocognitive recovery, represent a significant clinical challenge impacting patient recovery, quality of life, and healthcare systems [1]. Among the various risk factors, perioperative sleep disturbances are highly prevalent and constitute a significant independent risk factor for PND [2–4]. Postoperative sleep deprivation (PSD), in particular, has been closely associated with the development and exacerbation of PND, yet the precise mechanisms linking PSD to cognitive decline remain incompletely understood.

Mounting evidence underscores neuroinflammation as a pivotal central mechanism driving PND pathogenesis [5–8]. Both surgical trauma/anesthesia and perioperative sleep disturbances independently contribute to this neuroinflammatory cascade [9]. Surgical and anesthetic stimuli trigger a systemic inflammatory response, facilitating the entry of peripheral cytokines into the central nervous system (CNS). This influx activates microglia, the resident CNS immune cells, promoting their polarization towards a pro-inflammatory (M1) phenotype and the release of cytokines (e.g., TNF-α, IL-1β, IL-6) within the brain parenchyma, particularly the hippocampus—a region critical for learning and memory [5, 6, 10–12]. This microglia-driven neuroinflammation is widely recognized as a key contributor to synaptic dysfunction and cognitive decline in PND [10, 13, 14].

Concurrently, sleep deprivation (SD) itself is a potent inducer of neuroinflammation, exacerbating hippocampal cytokine expression and microglial activation [15–18]. Furthermore, SD compromises blood-brain barrier integrity and disrupts gut-brain axis communication, potentially amplifying peripheral-to-central inflammatory signaling and vice versa [15–18]. The intricate bidirectional relationship between sleep and brain health highlights the critical role of adequate sleep in maintaining CNS homeostasis, including metabolic waste clearance [19].

Recent breakthroughs have highlighted the meningeal lymphatic vessels (MLVs) as crucial regulators of CNS immune surveillance and waste clearance [20–23]. MLVs drain cerebrospinal fluid (CSF) and interstitial fluid (ISF), along with associated solutes (e.g., metabolic waste, inflammatory mediators), to the deep cervical lymph nodes (dCLNs) [22–26]. Impaired MLVs function is increasingly implicated in exacerbating neuroinflammation and pathology in various CNS disorders, including traumatic brain injury, Alzheimer’s disease models, and subarachnoid hemorrhage [20, 21, 27–30]. Critically, enhancing MLVs function (e.g., via vascular endothelial growth factor C, VEGF-C) can mitigate neuroinflammation and improve outcomes [20, 27, 28, 31]. Notably, sleep deprivation has been shown to impair brain molecular clearance in humans, an effect not fully reversible [32], and chronic sleep disruption can impair glymphatic function [33]. However, whether and how postoperative SD specifically impacts MLVs structure and function, thereby contributing to the exacerbated neuroinflammation and cognitive deficits observed in PND, remains a critical unanswered question.

Therefore, this study aimed to investigate the effects of postoperative sleep deprivation on MLVs morphology and drainage function in mice; Determine the role of MLVs dysfunction in mediating hippocampal neuroinflammation and cognitive impairment induced by PSD; and Assess whether therapeutic enhancement of MLVs function can alleviate these detrimental effects. We hypothesized that PSD impairs MLVs, leading to reduced clearance of inflammatory mediators, amplified hippocampal neuroinflammation, and cognitive dysfunction, and that augmenting MLVs function would mitigate these effects.

Materials and methods

Animals

Male C57BL/6JNifdc mice (age: 8–10 weeks; weight: 20–25 g) were sourced from Zhejiang Charles River Laboratory Animal Co., Ltd. Upon arrival at the Ningbo University Experimental Animal Center (SPF facility), mice underwent quarantine acclimation for 7 days in individually ventilated cages (IVC) under standardized conditions: 12-hour light/dark cycle (lights on 07:00–19:00), ambient temperature 22 ± 1 °C, relative humidity 50 ± 5%, and 15–20 air changes/hour. Animals received feed and water ad libitum throughout experiments. All procedures strictly complied with the Zhejiang Province Laboratory Animal Welfare Guidelines and were approved by the Animal Ethics Committee of the Medical Department of Ningbo University (Approval No.: NBU20210351).

Chemicals and antibodies

Ovalbumin conjugated to Alexa Fluor 647 (#A24785, Thermo Fisher Scientific, USA) served as the lymphatic tracer. Primary antibodies included: anti-LYVE-1-eFluor 488 (#AB1633415) and anti-LYVE-1-eFluor 660 (#AB10597449) from eBioscience (USA) for lymphatic endothelial labeling; anti-Iba-1 (#01919741, FUJIFILM Wako, Japan) for microglial detection; anti-IL-6 (#12912), anti-TNF-α (#11948) from Cell Signaling Technology (USA), and anti-IL-1β (#sc-32294, Santa Cruz, USA) for inflammatory cytokine detection; with anti-β-actin (#20536, Proteintech, China) as loading control. Species-matched secondary antibodies comprised Alexa Fluor 488-conjugated goat anti-rabbit IgG (#AS053, ABclonal, China) for fluorescence imaging, plus HRP-conjugated goat anti-rabbit (#A0208) and anti-mouse (#A0216) antibodies from Beyotime (China) for immunoblotting. All antibodies underwent validation via positive control tissues before experimental use.

Recombinant adeno-associated virus vector construction

Recombinant adeno-associated virus (AAV) vectors expressing vascular endothelial growth factor C (VEGF-C) or enhanced green fluorescent protein (EGFP) under CMV promoters were commercially generated (Shanghai Heyuan Biotechnology, China). The mouse VEGF-C coding sequence (NCBI Reference: NM_009506.2) was PCR-amplified and cloned into the pcAAV-CMV-MCS-WPRE backbone to generate pcAAV-CMV-VEGF-C-WPRE, while the control construct pcAAV-CMV-EGFP-WPRE contained the EGFP reporter. Both plasmids were packaged into AAV1 capsids, with viral titers quantified at 1 × 10¹² genome copies/mL. Viruses were transported on dry ice and stored at -80 °C in light-protected vials, maintaining stability for six months. For intracisternal injections, aliquots were thawed rapidly at 37 °C immediately before use.

Establishment of PSD animal model

Mice underwent tibia fracture surgery (TFS) under isoflurane anesthesia (induced with 3–4%, maintained at 1.5-2% in 100% oxygen at 2 L/min, adjusted based on respiratory rate and pedal reflex). Surgical instruments were sterilized. Under aseptic conditions, the right lower leg was shaved and disinfected with iodine. A ~ 3 mm skin incision was made parallel to the tibia in the upper third to expose the tibial plateau. A hole was drilled into the plateau with a 25-gauge needle. A 0.36 mm diameter intramedullary pin was inserted into the medullary cavity to a depth of ~ 15 mm, cut flush with the plateau, and secured. The middle segment of the tibia was exposed via blunt dissection and fractured using bone cutters, avoiding neurovascular damage. The wound was sutured (4 − 0 silk) and topically treated with erythromycin ointment. Control mice received anesthesia and skin incision without fracture. Mice were monitored postoperatively; those showing no weight-bearing on the operated limb 24 h later were excluded. Twenty-four hours post-surgery, mice assigned to sleep deprivation (SD) or TFS + SD groups were placed on a treadmill (SA101, SansBio, China) with free access to food and water. The treadmill was programmed for intermittent sleep deprivation: 30 s of movement at 3 m/minute followed by 90 s of rest (duty cycle: 25% on), repeating continuously for 72 h, achieving approximately 30 deprivation bouts per hour.

Cisterna magna injection

Mice were anesthetized and secured in a stereotactic frame, with the head-neck angle maintained at approximately 120°. The dorsal midline skin between the ears was shaved and aseptically prepared with iodine. A 3-mm midline incision was surgically created using fine scissors. Subcutaneous tissues were gently dissected with ophthalmic forceps to expose the cisterna magna, identified as an inverted Y-shaped triangular area. A 33-gauge needle connected to a 10-µl Hamilton microsyringe via polyethylene tubing was primed with 2–5 µl of experimental reagent ( [2 µl (for AAV) or 5 µl (for OVA)]). The needle was inserted into the cisterna magna at a 45° angle to a depth of 1 mm, and the reagent was infused at 2 µl/min. Post-injection, the needle remained in situ for 2 min to minimize backflow. Tribromoethanol (Avertin) was used for this brief microsurgical procedures. The solution was freshly prepared before each use, and no adverse effects were observed.

Deep cervical lymphatic vessel ligation

Mice were anesthetized with tribromoethanol (125 mg/kg), and the ventral neck skin was aseptically prepared with povidone-iodine. A 3-mm midline incision was made 5 mm cranial to the clavicle. Under surgical stereomicroscope, subcutaneous tissues were dissected to expose the right dCLNs at the junction of the sternocleidomastoid muscle and trachea. Afferent lymphatic vessels (dCLVs) draining into the rostral pole of dCLNs were identified and ligated with 10 − 0 non-absorbable nylon sutures. Extreme care was taken to avoid adjacent neurovascular structures (vagus nerve, carotid artery). The incision was closed with 6 − 0 silk sutures and disinfected. Postoperatively, mice were monitored for 24 h for hemorrhage, infection, or neurological deficits. The mean operative duration was approximately 8 min, with no adverse events observed.

Open field test (OFT)

Mice were habituated to the testing environment for 1 h under dim illumination prior to behavioral assessment. Spontaneous exploratory behavior was evaluated in a white polyvinyl chloride arena (50 × 50 × 30 cm) under uniform overhead lighting. Each mouse was initiated from a corner position facing the arena wall, and locomotor activity was recorded for 5 min using a ceiling-mounted digital camera. Video analysis was performed using Shanghai Xinruan Behavioral Tracking Software to quantify total distance traveled (cm) and immobility time (s). Between subjects, the apparatus was thoroughly cleansed with 70% ethanol and dried with lint-free wipes to eliminate residual olfactory cues.

Y-Maze test (YMT)

Following identical habituation protocols, spatial working memory was assessed in a Y-shaped maze (arm dimensions: 50 cm length × 20 cm height; inter-arm angle: 120°) under standardized illumination. Mice were introduced at the distal terminus of a single arm oriented toward the central junction. Unrestricted exploration was video-recorded for 8 min using an overhead tracking system. Behavioral parameters were extracted via Shanghai Xinruan Software, including: (1) Total arm entries, quantified when all four limbs crossed the arm threshold; and (2) Spontaneous alternation rate, calculated using the formula: Alternation rate= (Total alternations / Maximum alternations) × 100%, where a complete triad constitutes consecutive entries into three distinct arms. The maze underwent systematic decontamination between trials with 70% ethanol and 5-minute air-drying cycles to prevent olfactory interference, with experimenters absent during recording sessions to eliminate observational bias.

Contextual fear conditioning (CFC)

Contextual fear memory assessment was conducted in a standardized conditioning chamber (32 × 32 × 34 cm). Following 1-hour environmental habituation, mice underwent training consisting of: 2-minute free exploration, two 0.75 mA footshocks (2-second duration, 30-second inter-stimulus interval), and 30-second post-shock observation. Chamber surfaces were systematically decontaminated between subjects with 70% ethanol and lint-free wipes to eliminate odor cues. After 24 h, memory retention was tested during a 3-minute free exploration session without shocks. Freezing behavior (complete immobility except respiration) was quantified using Shanghai Xinruan Software.

Object recognition test (ORT)

Object recognition memory was evaluated in an open field arena (50 × 50 × 50 cm). After 1-hour habituation, mice underwent training with two identical rectangular objects (Object A1/A2; 4 × 4 × 6 cm) positioned 10 cm from adjacent walls. Following 5-minute exploration, mice returned to home cages. After 24 h, one familiar object was replaced with a novel spherical object (Object B; Ø6 cm) in identical spatial configuration. Exploration was video-recorded for 5 min and analyzed using Xinruan Software (Shanghai Xinruan Co., Ltd., China), with exploration defined as the mouse touching the objects with its whiskers or paws excluding climbing/circling. Recognition index (RI) was calculated as: RI = [Exploration time Novel Object / (Exploration time Novel + Familiar Objects)] × 100%. Chamber decontamination employed 70% ethanol with 5-minute drying intervals between trials.

Tissue harvesting and processing

Mice were deeply anesthetized with tribromoethanol (125 mg/kg, Aladdin, Shanghai, China) and secured in a supine position. The dCLNs were surgically excised under stereomicroscopic guidance, preserving adjacent neurovascular structures. Following dCLN removal, transcardial perfusion was performed with 20 ml ice-cold 0.9% saline via the left ventricle. Craniotomy was executed to extract whole brains, which were immersion-fixed in 4 °C tissue fixation solution (4% paraformaldehyde in PBS) for 24 h concurrently with skull caps. Brains and dCLNs were dehydrated in 15% and 30% sucrose solutions before embedding in optimal cutting temperature compound. Coronal Sect.  (30 μm thickness) were prepared using a cryostat (Leica CM1950) at -20 °C chamber temperature.

Fixed skull caps were decalcified in EDTA (0.5 M, pH 7.4) for 48 h, rinsed in PBS, and subjected to micro-dissection under 10× magnification. The calvarial inner surface was meticulously cleared of connective tissue using ophthalmic forceps, with intact meninges peeled from sagittal/lambdoid sutures and stored in PBS at 4 °C.

Immunofluorescence staining and quantitative analysis

Tissue sections and whole-mount meninges underwent permeabilization in PBST (0.1% Triton X-100/PBS) for 30 min, followed by blocking with 1% bovine serum albumin at room temperature for 1.5 h. Primary antibody incubation was performed overnight at 4 °C using the following reagents: anti-LYVE-1-eFluor 660 (1:200, eBioscience #AB10597449) for lymphatic endothelial labeling, anti-LYVE-1-eFluor 488 (1:200, eBioscience #AB1633415) for dual-channel lymphatic visualization, and anti-Iba-1 (1:200, FUJIFILM Wako #01919741) for microglial detection. After three 10-minute PBST washes, sections were incubated with Alexa Fluor 488-conjugated goat anti-rabbit IgG secondary antibody (1:400, Abclonal #AS053) for 2 h. Nuclear counterstaining employed DAPI (1 µg/ml) prior to mounting in ProLong Diamond antifade medium (ThermoFisher) to preserve fluorescence signal integrity.

Confocal imaging was executed on a Leica SP8 system equipped with white-light laser excitation and hybrid detectors (HyD), implementing standardized acquisition protocols across experimental groups: MLVs morphological assessment employed a 20×/0.75 multi-immersion objective at 512 × 512 resolution with 10 μm z-steps. Quantitative image analysis encompassed three principal metrics: Microglial activation indexed by mean Iba-1 fluorescence intensity in hippocampal DG, CA1, and CA3 subfields, derived from five anatomically matched sections per subject; Lymphatic drainage efficiency calculated as percentage area of OVA-647⁺/LYVE-1⁺ co-localization in dCLN sections; and MLVs morphometrics including vessel diameter (30 measurements per transverse sinus region) and LYVE-1⁺ coverage area (%), both computed from maximum intensity projections. Sholl analysis was performed on confocal images of brain sections immunostained with Iba1 antibody. Microglial morphology was quantitatively assessed using ImageJ software equipped with the Sholl Analysis plugin. The segmented line tool was employed to manually trace Iba1-positive microglial processes. Concentric circles were subsequently generated by the plugin centered on the cell soma, with a terminating radius of 60 μm and radial intervals of 2 μm. The number of intersections between these concentric circles and microglial processes was automatically calculated at each radial distance. All analytical procedures utilized ImageJ software (NIH, USA).

Western Blot (WB)

Hippocampal tissues were homogenized in RIPA lysis buffer (Beyotime #P0013B) containing protease inhibitors, followed by centrifugation at 13,500 × g (4 °C, 30 min). Protein concentrations were determined via BCA assay (Pierce™), with 30 µg lysates denatured in Laemmli buffer at 100 °C. Samples underwent SDS-PAGE (12.5% gels) and wet transfer to PVDF membranes (Millipore). After blocking with 5% skim milk/TBST, membranes were probed overnight at 4 °C with primary antibodies against: IL-1β (1:200, Santa Cruz #sc-32294), IL-6 (1:1000, CST #12912), TNF-α (1:1000, CST #11948), and β-actin (1:2000, Proteintech #20536). Species-matched IRDye^® secondary antibodies (LI-COR) enabled detection on an Odyssey Imaging System. Stripping/reprobing used Epizyme buffer (#PS107) for multiplex analysis. Band intensities were quantified using ImageJ with β-actin normalization.

Meningeal tissue RNA sequencing and bioinformatic analysis

The meninges were carefully dissected, placed in sterile, RNase-free tubes containing RNA stabilizing solution, snap-frozen on dry ice, and shipped to Shanghai Majorbio Bio-pharm Technology Co., Ltd. on the same day. Total RNA was extracted from tissue samples, and mRNA was enriched using Oligo dT. The enriched mRNA was fragmented to ~ 300 bp, reverse-transcribed into first-strand cDNA using random primers, followed by second-strand synthesis. The resulting double-stranded cDNA was end-repaired, A-tailed, ligated with adapters, size-selected, PCR-amplified, and purified to generate the final library. Sequencing was performed on the NovaSeq X Plus platform.

Statistical analysis

All quantitative data were analyzed using GraphPad Prism 9.1.1 software (GraphPad Software, USA). Continuous variables are expressed as mean ± standard error of the mean (SEM). For multi-group comparisons (e.g., Control/TFS/SD/TFS + SD cohorts), two-way ANOVA was employed followed by tukey’s post hoc test for pairwise analyses. Therapeutic intervention studies (e.g., VEGF-C versus EGFP controls) employed one-way analysis of variance (ANOVA) followed by Tukey’s post-hoc test for multiple group comparisons. Statistical significance was defined at P < 0.05.

Results

PSD following surgery induces cognitive decline in mice

To assess the impact of PSD on cognitive function, mice underwent behavioral testing (Fig. 1A). Critically, PSD (TFS + SD) induced significant impairments in spatial working memory and contextual fear memory. OFT revealed no differences in average speed (Fig. 1C) or total distance traveled (Fig. 1D) among groups, indicating comparable locomotor activity. ORT also showed no significant differences (Fig. 1F). However, in the YMT, TFS + SD mice exhibited a significantly reduced spontaneous alternation rate compared to Control and TFS groups (P < 0.01; Fig. 1H), indicating impaired spatial working memory. No differences were observed in total arm entries (Fig. 1G). Furthermore, in the CFC, TFS + SD mice displayed significantly shorter freezing times compared to Control and TFS groups (P < 0.01; Fig. 1I), reflecting impaired hippocampus-dependent contextual fear memory.

Fig. 1.

PSD induced cognitive dysfunction in mice. A Experimental timeline of behavioral tests. B Representative track images of OFT for mice in each group. C Average speed in OFT. D Total distance traveled in OFT. E Representative track images of ORT for mice in each group. F Recognition index in ORT. G Total arm entries in YMT. H Spontaneous alternation rate (%) in YMT. I Freezing time (%) in CFC. Data: Mean ± SEM; n = 9–10 mice/group. *P < 0.05, **P < 0.01 (two-way ANOVA with tukey’s post hoc). Abbreviations: PSD, Postoperative sleep deprivation; OFT, Open field test; YMT, Y-Maze test; ORT, Object recognition test; CFC, Contextual fear conditioning

PSD-induced transcriptomic alterations in mouse meningeal tissues

To gain molecular insights into the meningeal pathology underlying PSD-induced cognitive decline, we performed bulk RNA sequencing on whole-mount meningeal tissues from Control and TFS + SD mice. Transcriptomic analysis revealed 209 significantly down-regulated genes in TFS + SD MLVs versus controls (Fig. 2A, B). GO enrichment identified prominent reductions in response to radiation, transmembrane signaling receptor activity, and positive regulation of cell activation (Fig. 2C), indicating broad disruption of biological processes and molecular functions. KEGG annotation showed down-regulated genes linked to sensory/endocrine systems and signal transduction (Fig. 2D), suggesting impaired inter-system communication. Reactome analysis confirmed diminished expression of immune-system and signal-transduction genes (Fig. 2E), implying compromised immunological competence and intracellular signaling in MLVs.

Fig. 2.

Comprehensive transcriptome sequencing results analysis chart. A Bar graph summarizing differential gene-expression changes in meningeal tissues. B Clustered heatmap of differentially expressed genes in meningeal tissues. C GO enrichment analysis of differentially expressed genes in meningeal tissues. D KEGG pathway classification of differential genes in meningeal tissues. E Reactome pathway classification of differential genes in meningeal tissues

These unbiased transcriptomic data reveal that a profound dysregulation of immune responses is central to the pathology of TFS + SD. This finding serves as a critical molecular rationale, directly motivating our subsequent investigation into the accumulation of specific pro-inflammatory cytokines and the functional integrity of the meningeal lymphatic vessels responsible for their clearance.

PSD exacerbates hippocampal neuroinflammation in mice

Immunofluorescence staining demonstrated robust microglial activation (Iba1 + intensity) in hippocampal CA1, CA3 and DG area of TFS + SD mice, exceeding levels observed in control groups (P < 0.0001, Fig. 3B-E). We also conducted Sholl analysis on individual microglia. Our results indicated that the total number of intersections was significantly higher in the TFS + SD group compared to the control group (P < 0.001, Fig. 3F - H).

Fig. 3.

PSD leads to upregulation of IL-1β, IL-6, and TNF-α levels and increased microglial activation in mouse hippocampi. A Experimental scheme of immunofluorescence staining and WB for hippocampus. B Immunofluorescence of Iba1-labeled microglia (green) in CA1, CA3, DG. Scale bar: 100 μm. C-E Iba1 intensity in CA1, CA3, DG. F-H The number of process intersections at various radial distances from the microglial soma in CA1, CA3, DG. I Representative Western blot images of IL-1β, IL-6, TNF-α, and β-actin in the hippocampus on the fourth day following model establishment. J-L The relative expression levels of hippocampus IL-1β (J), IL-6 (K), TNF-α (L). M Representative Western blot images of IL-1β, IL-6, TNF-α, and β-actin in the hippocampus on the fifth day following model establishment. N-P The relative expression levels of hippocampus IL-1β (N), IL-6 (O), TNF-α (P). Data are presented as Mean ± SEM. C-H n = 5; (J-L, N-P) n = 4 mice/group. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 (one-way ANOVA (F-H) or two-way ANOVA (C-E, J-L, N-P) with tukey’s post hoc). Abbreviations: PSD, Postoperative sleep deprivation

Western blot analysis further revealed increases in IL-1β, IL-6, and TNF-α expression in hippocampi of TFS + SD mice versus control (P < 0.01, Fig. 3I-P). At Day 4 (immediately after SD), there was no significant difference between the SD and TFS + SD groups in IL-1β or IL-6 expression (Fig. 3J-L). However, cytokine expression in the TFS + SD group was significant higher than in the SD group at Day 5 (following a 24-hour sleep recovery period) (P < 0.05 or P < 0.01; Fig. 3N-P). These findings indicate that the combination of surgical trauma and sleep deprivation not only amplifies the acute neuroinflammatory response but also prolongs its duration, even after sleep recovery, underscoring a sustained contribution of surgery to hippocampal inflammation.

PSD impairs MLVs structure and drainage function

To investigate the impact of PSD on MLVs integrity, meningeal tissues were harvested for whole-mount Lyve-1 immunofluorescence staining. Quantitative analysis revealed that MLVs in PSD mice exhibited significantly reduced vessel diameter and decreased coverage area (Fig. 4C, D), indicating structural compromise (Fig. 4B).

Fig. 4.

Structural and functional deficits in meningeal lymphatics after PSD. A Experimental scheme of immunofluorescence staining for MLVs. B Whole-mount Lyve-1+ (MLVs; red). Scale bars: 1 mm (overview), 100 μm (transverse sinus inset). C MLVs diameter quantification. D MLV coverage area (%). E Experimental scheme of immunofluorescence staining for dCLNs. F Representative dCLN sections: OVA-647 (red), Lyve-1 (green). Scale bar: 200 μm. G Schematic diagram of cisterna magna injection. H OVA-647 fluorescence area (%) in dCLNs 2 h post-cisterna magna injection. I Experimental scheme of WB for ligated vs. non-ligated dCLNs dCLNs. J Representative Western blot images of IL-1β, IL-6, TNF-α, and β-actin in dCLNs on the fourth day following model establishment. K-M The relative expression levels of dCLNs IL-1β (K), IL-6 (L), TNF-α (M). N Representative Western blot images of IL-1β, IL-6, TNF-α, and β-actin in dCLNs on the fifth day following model establishment. O-Q The relative expression levels of dCLNs IL-1β (O), IL-6 (P), TNF-α (Q). R Schematic diagram of deep cervical lymphatic vessel ligation. S Representative Western blot images of IL-1β, IL-6, TNF-α, and β-actin in dCLNs from both sides after unilateral lymphatic ligation. T-V The relative expression levels of IL-1β (T), IL-6 (U), TNF-α (V) in ligated vs. non-ligated dCLNs. Data are presented as Mean ± SEM.; (C, D) n = 6; (H, K-M, O-Q, T-V) n = 4 mice. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 (C, D, H, K-M, O-Q: two-way ANOVA with tukey’s post hoc; T-V one-way ANOVA with tukey’s post hoc). Abbreviations: PSD, Postoperative sleep deprivation; MLVs, Meningeal lymphatic vessels; dCLN, deep cervical lymph nodes

To assess functional drainage, 5 µl of AF647-conjugated ovalbumin (OVA-647) was injected into the cisterna magna (Fig. 4E, G). dCLNs were harvested 2 h post-injection, sectioned, and co-stained for Lyve-1. Confocal imaging demonstrated significantly lower OVA-647⁺ area in dCLNs of TFS + SD mice versus Control/TFS/SD groups (Fig. 4F, H), confirming impaired solute clearance.

To verify the CNS origin of these cytokines, unilateral dCLN afferent lymphatic ligation was performed (Fig. 4I, R), followed by PSD 12 h thereafter. dCLNs from the non-ligated side exhibited higher cytokine levels than ligated counterparts (Fig. 4S–V), confirming that dCLN cytokines originate predominantly from brain-meningeal drainage.

We further evaluated inflammatory cytokine drainage to dCLNs at both day 4 and day 5. At day 4 and 5, protein levels of IL-1β, IL-6, and TNF-α were markedly reduced in TFS + SD dCLNs compared to TFS groups (Fig. 4J–Q), reflecting impaired clearance of CNS-derived cytokines. Specifically, the levels of IL-1β, IL-6, and TNF-α in dCLNs were significantly lower in the TFS + SD group compared to the SD group at Day 5 (Fig. 4N–Q). This persistent drainage deficit in TFS + SD mice suggests that surgery exacerbates the impairment of MLV function, delaying recovery of clearance capacity.

Enhancing MLVs via VEGF-C overexpression alleviates cognitive dysfunction induced by PSD

To determine if rescuing MLV function could mitigate cognitive deficits, we overexpressed VEGF-C intracisternally using AAV (Fig. 5A). Following PSD, mice overexpressing VEGF-C (TFS + SD+VEGF-C) exhibited significant improvement in cognitive performance compared to those receiving control virus (TFS + SD+EGFP). Locomotor activity (OFT total distance, average speed; Fig. 5G, H) and exploratory drive (YMT total arm entries; Fig. 5D) remained comparable across all groups. Importantly, TFS + SD+VEGF-C mice showed a significantly higher spontaneous alternation rate in the YMT than TFS + SD+EGFP mice (P < 0.05; Fig. 5C). Furthermore, freezing time in the CFC test was significantly increased in the TFS + SD+VEGF-C group compared to the TFS + SD group (P < 0.05; Fig. 5E).

Fig. 5.

PSD-induced cognitive decline was rescued by VEGF-C overexpression. A Schematic diagram of cisterna magna injection and post-AAV injection behavioral timeline. B YMT spontaneous alternation rate (%). C YMT total arm entries. D CFC freezing time (%). E Representative track images of OFT for mice in each group. F OFT total distance. G OFT average speed. Data are presented as Mean ± SEM; n = 8–10 mice/group. *P < 0.05, **P < 0.01, ***P < 0.001 (one-way ANOVA with tukey’s post hoc). Abbreviations: VEGF-C, Vascular endothelial growth factor C. YMT, Y-Maze test; CFC, Contextual fear conditioning; OFT, Open field test

Overexpression of VEGF-C alleviates PSD-induced morphology alteration and drainage function impairment of MLVs

We confirmed that AAV-VEGF-C effectively targeted and augmented MLVs. Immunofluorescence revealed significantly increased MLV diameter and coverage in TFS + SD+VEGF-C mice compared to TFS + SD+EGFP controls (Diameter: P < 0.001; Coverage: P < 0.01; Fig. 6A, C, D). EGFP expression confirmed viral transduction around MLVs (Fig. 6B). Functionally, OVA-647 drainage to dCLNs was significantly enhanced in the TFS + SD+VEGF-C group compared to the TFS + SD+EGFP group (P < 0.05; Fig. 6E, F). Consistent with improved drainage capability, levels of inflammatory cytokines (IL-1β, IL-6, TNF-α) in dCLNs were significantly higher in the TFS + SD+VEGF-C group than in the TFS + SD+EGFP group (all P < 0.05; Fig. 6H-K), indicating more efficient clearance of these mediators from the CNS.

Fig. 6.

VEGF-C overexpression improves meningeal lymphatic morphology and drainage function. A Representative immunofluorescence images of MLVs in the whole meninges (red). Scale bars: 1 mm (overview), 100 μm (inset). B EGFP expression (green) adjacent to MLVs in TFS + SD+VEGF-C mice. Scale bars: 1 mm (left), 100 μm (right). C MLV diameter. D MLV coverage area (%). E OVA-647 drainage in dCLNs. Scale bar: 200 μm. F OVA-647 area (%) in dCLNs. G Experimental scheme of cisterna magna injection. H Representative Western blot images of IL-1β, IL-6, TNF-α, and β-actin in dCLNs. I-K The relative expression levels of dCLNs IL-1β (I), IL-6 (J), TNF-α (K). Data are presented as Mean ± SEM; (E, F) n = 6; (G-J) n = 4 mice/group. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 (one-way ANOVA with tukey’s post hoc). Abbreviations: VEGF-C, Vascular endothelial growth factor C; + MLVs, Meningeal lymphatic vessels

VEGF-C overexpression attenuates hippocampal neuroinflammation induced by PSD

Finally, we examined the impact of improved MLV function on hippocampal neuroinflammation. Immunofluorescence analysis showing significantly attenuated microglial activation in the CA1, CA3, and DG subfields of TFS + SD+VEGF-C mice compared to TFS + SD+EGFP mice (CA1: P < 0.01; CA3: P < 0.01; DG: P < 0.05; Fig. 7B-E). Additionally, Sholl analysis was performed on individual microglia (Fig. 7F - H). The results indicated that the total number of intersections was significantly reduced in the TFS + SD+VEGF-C group compared with the TFS + SD+EGFP group (CA1: P < 0.001; CA3: P < 0.0001; DG: P < 0.01; Fig. 7F-H). Western blot analysis further demonstrated that hippocampal levels of IL-1β, IL-6, and TNF-α were significantly reduced in TFS + SD+VEGF-C mice compared to TFS + SD+EGFP controls (all P < 0.05; Fig. 7J-L).

Fig. 7.

Attenuation of neuroinflammation by meningeal lymphatic potentiation. A Experimental scheme of immunofluorescence staining and WB for hippocampus. B Iba1-labeled microglia in hippocampal subfields. Scale bar: 100 μm. C-E Iba1 intensity in CA1, CA3 and DG, respectively. F The number of process intersections at various radial distances from the microglial soma in CA1. G The number of process intersections at various radial distances from the microglial soma in CA3. H The number of process intersections at various radial distances from the microglial soma in DG. I Western blot analysis was used to detect protein expression of IL-1β, IL-6 and TNF-α in each fraction, β-actin: marker. J-L The relative expression levels of hippocampal IL-1β (J), IL-6 (K), TNF-α (L). Data are presented as Mean ± SEM. C-E n = 5; (G-I) n = 4 mice/group. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 (one-way ANOVA with tukey’s post hoc)

Discussion

This study provides new evidence that PSD disrupts MLVs’ structure and function, driving hippocampal neuroinflammation and cognitive impairment in mice. Critically, enhancing MLV function via intracisternal AAV-VEGF-C delivery rescues lymphatic drainage, attenuates neuroinflammatory responses, and improves cognitive performance. These findings establish MLVs as pivotal regulators of PND-related cognition decline and identify them as a potential therapeutic targets (Fig. 8).

Fig. 8.

Integrated mechanism of PSD-induced cognitive decline and therapeutic rescue via meningeal lymphatic enhancement

There is a significant reduction in MLVs diameter and dural coverage after the combined insult, whereas isolated surgery or sleep deprivation left the lymphatic architecture largely intact. Functional tracing with intracisternally injected fluorescent ovalbumin showed delayed transfer to dCLNs, and unilateral ligation of cervical lymphatics confirmed that cytokines detected in dCLNs originated predominantly from brain–meningeal outflow because protein levels on the ligated side were markedly lower than on the intact side [34–36]. These observations resonate with models of traumatic brain injury, Alzheimer’s disease and subarachnoid haemorrhage in which impaired MLVs drainage augments parenchymal inflammation and hastens functional decline [20, 24–28, 37, 38]. The fact that lymphatic dysfunction emerged only when metabolic load and inflammatory signalling converged suggests that compensatory capacity in young animals suffices for a single stressor but fails when clearance demand outstrips anatomical reserve. Our analysis focused on the dorsal MLVs, which are highly plastic and drain the hippocampal region; whether PSD similarly affects other CNS-associated lymphatic routes, such as those at the skull base or nasopharynx, remains an important avenue for future research.

A critical insight from our temporal analysis is the distinct contribution of surgical trauma versus sleep deprivation to the pathology of PND. While both the SD and TFS + SD groups exhibited comparable severe cognitive deficits and peak inflammation immediately after the 72-hour stress protocol (Day 4), their recovery trajectories diverged significantly. By extending our observation to Day 5 (24 h of recovery), we found that the SD group exhibited a significant trend towards inflammatory resolution and functional lymphatic recovery. In contrast, the TFS + SD group displayed a “failure to resolve” phenotype, characterized by persistently elevated hippocampal cytokines and severely impaired lymphatic drainage. This dissociation highlights that while both surgical trauma and sleep deprivation challenge the meningeal lymphatic system, the combination induces a profound structural compromise that overwhelms the brain’s resilience mechanisms. Specifically, the superimposition of surgical stress exacerbates the lymphatic dysfunction initiated by sleep deprivation, transforming a potentially reversible physiological strain into a persistent pathological state with impaired clearance capacity.

Transcriptomic profiling revealed distinct clustering of differentially expressed genes in PSD meningeal tissues, indicating state-dependent functional specificity. GO enrichment demonstrated significant down-regulation of immune-response pathways, suggesting altered inflammatory regulation. KEGG and Reactome analyses further highlighted disruptions in circadian-rhythm pathways (e.g., clock gene networks) and signal transduction cascades, aligning with sleep deprivation-induced biological clock dysregulation. These transcriptional changes likely impair MLV responsiveness to physiological cues (e.g., neurotransmitters, hormones), exacerbating drainage deficiency.

To interrogate causality, we augmented meningeal lymphangiogenesis through intracisternal delivery of an adeno-associated viral vector encoding vascular endothelial growth factor-C. Four weeks later, VEGF-C over-expression enlarged MLVs lumen diameter, restored dural coverage and normalised ovalbumin trafficking to dCLNs. The redistribution of inflammatory mediators was equally striking: dCLN cytokine content increased while hippocampal levels diminished, paralleled by attenuation of microglial activation and recovery of Y-maze alternation and contextual freezing. These results extend seminal work by da Mesquita and colleagues, who showed that VEGF-C-mediated lymphatic expansion facilitates clearance of amyloid-β and potentiates immunotherapy in Alzheimer models [27, 28], and they dovetail with magnetic-resonance evidence for a patent skull-base lymphatic conduit in humans [24, 25]. Collectively, our findings place MLVs at the centre of a pathological loop in which post-operative sleep loss throttles waste clearance, allowing peripherally triggered cytokines to accumulate in the parenchyma, perpetuate microglial inflammation and impair memory. In addition, the observed MLV impairment following PSD likely reflects a vulnerability specific to the CNS-lymphatic interface, possibly due to disrupted circadian CSF dynamics and direct meningeal inflammation, rather than a systemic lymphatic failure. Future comparative studies of peripheral and meningeal lymphatic health in this model would be valuable to confirm this specificity.

The study also draws attention to the intimate coupling between the glymphatic network and dural lymphatics. Deep, non-fragmented sleep polarises astrocytic aquaporin-4 and thereby accelerates glymphatic influx, whereas chronic sleep fragmentation reduces paravascular flow and favours retention of neurotoxic proteins such as amyloid-β and phosphorylated tau [33, 39]. Although we did not directly quantify glymphatic flux, it is plausible that the observed MLVs deficits co-occurred with impaired glymphatic transport, creating a two-node bottleneck for solute clearance. Future work that combines two-photon imaging of peri-vascular spaces with real-time lymphatic flowometry is warranted to disentangle the relative contributions of these pathways and to determine whether targeted stimulation of glymphatic influx alone suffices to restore lymphatic patency under post-operative conditions.

Previous studies utilizing the tibial fracture model have reported cognitive deficits primarily in aged mice [40, 41], where meningeal lymphatic function is already compromised [42]. In contrast, our study employed young adult mice, which exhibited resilience to TFS alone, showing no significant cognitive impairment. This resilience in young mice provided a unique window to isolate the specific contribution of sleep deprivation. Our data demonstrate that even in a resilient young brain, the superimposition of sleep disturbance upon surgical stress is sufficient to trigger lymphatic dysfunction and subsequent cognitive decline, highlighting SD as a potent, independent risk factor regardless of age.

Several limitations should be acknowledged. First, we used young male mice to minimize variability; whether VEGF-C confers similar benefits in aged or female animals—where baseline MLV function is already compromised—remains to be determined [28, 43]. Second, while our ligation experiment supports a drainage-centered mechanism, VEGF-C’s pleiotropic vascular and immunomodulatory effects cannot be fully excluded. More selective tools for modulating lymphatic function are needed to address this.Third, while our model captures the acute phase of PSD-induced decline, future studies employing extended recovery periods are warranted to determine whether MLV dysfunction also underlies prolonged or delayed PND, and whether therapeutic intervention timing influences long-term cognitive outcomes. Fourth, future investigations profiling the immune cell landscape in cervical lymph nodes following enhanced CNS drainage could reveal how peripheral immune surveillance adapts to cleared neuroinflammatory signals. Fifth, we focused on microglial activation as the primary cellular link between cytokine dysregulation and cognitive impairment. While astrocytes also contribute to neuroinflammation and may interact with the glymphatic system, their role in this PSD model warrants future investigation. Sixth, the transcriptomic analysis was performed on whole meningeal tissue comparing only Control and TFS + SD groups. While this identified pathways relevant to our hypothesis, it does not distinguish the specific contributions of surgery versus sleep deprivation. Future studies with intermediate control groups will be needed for such attribution. Similarly, our VEGF-C intervention was applied in the context of the combined TFS + SD insult, demonstrating its therapeutic potential in a model integrating both key risk factors. Future studies employing VEGF-C in models of SD alone or TFS alone will be valuable to determine whether lymphatic enhancement broadly rescues sleep-loss pathology or specifically disrupts the synergistic interaction between surgery and sleep deprivation. Finally, our behavioural battery focused on hippocampus-dependent tasks; evaluation of executive function and long-term outcomes will be necessary to establish whether lymphatic modulation yields durable benefits across cognitive domains.

Several methodological considerations warrant mention. First, while direct EEG/EMG recording would provide detailed sleep architecture data, we opted for a non-invasive treadmill SD protocol to preclude any potential damage to the meningeal lymphatics from cranial electrode implantation. This ensured the reliability of our primary structural and functional assessments of MLVs. Future studies employing non-invasive sleep monitoring techniques (e.g., piezoelectric systems) could concurrently characterize sleep disturbances and lymphatic function. In addition, an interesting secondary question is whether improving CNS clearance via VEGF-C can subsequently normalize sleep architecture disrupted by PSD. While our study was not designed to assess sleep, this represents a compelling feedback loop for future investigation.

Despite these caveats, the present work delivers two clinically relevant messages. First, sleep management must be viewed as an obligatory component of neuroprotection in the peri-operative period, not only because adequate sleep reduces systemic stress but also because it safeguards central clearance pathways. Second, MLVs constitute an actionable therapeutic target: minimally invasive cervical lymphatic–venous anastomosis, low-dose lymphangiogenic agonists such as VEGF-C or mTOR-sparing rapalogs and neuromodulatory techniques synchronised with deep-sleep stages could be explored as adjuncts to current anti-inflammatory strategies. By illuminating the sleep–lymphatic–inflammation nexus, our study paves the way for translational approaches that aim to preserve brain homeostasis when it is most vulnerable.

Pathological Cascade

Postoperative sleep deprivation (PSD) triggers structural regression of meningeal lymphatic vessels (MLVs), characterized by reduced diameter and dural coverage (Fig. 4), concomitant with transcriptomic down-regulation of immune-response, circadian-rhythm, and signal-transduction genes (Fig. 2). This impairs clearance of hippocampal cytokines (IL-1β, IL-6, TNF-α) to deep cervical lymph nodes (dCLNs), leading to neuroinflammatory accumulation (Fig. 2), microglial hyperactivation, and cognitive deficits in spatial/contextual memory (Fig. 1).

Therapeutic Intervention

Intracisternal AAV-VEGF-C delivery rescues MLV morphology and drainage function (Fig. 6), restoring cytokine efflux to dCLNs. This attenuates hippocampal neuroinflammation (Fig. 7) and reverses cognitive impairment (Fig. 5), demonstrating that MLV potentiation disrupts the PSD-induced neuroinflammatory cycle.

Conclusion

In summary, PSD inflicts a dual injury comprising exacerbated hippocampal inflammation and meningeal lymphatic collapse, culminating in selective memory deficits. Genetic enhancement of MLVs function via VEGF-C reinstates lymphatic drainage, curtails neuroinflammation and rescues cognition. By incorporating the meningeal lymphatic system into current models of PND, we reveal a previously underappreciated bottleneck in peri-operative brain health and highlight lymphatic potentiation as a promising avenue for future clinical intervention.

Authors’ contributions

**Binkang Huang** : Formal analysis, Investigation, Visualization, Writing – original draft. **Wei Tu** : Formal analysis, Investigation, Visualization, Writing – original draft. **Han Zhou** : Investigation, Formal analysis. **Lan Mo** : Investigation. **Qiubo Yang** : Investigation. **Yilong Sheng** : Investigation. **Daofan Sun** : Investigation. **Bo Meng** : Investigation. **Zhexi Chi** : Investigation. **Yingying Lv** : Investigation. **Rongjun Liu** : Investigation. **Xiuzhong Xing** : Investigation. **Hui Yuan** : Investigation. **Jing Yang** : Funding acquisition Writing – review & editing. **Fengdeng Hu** : Writing – review & editing. **Guo Chen** : Writing – review & editing. **Junping Chen** : Supervision, Writing – review & editing. **Xiaowei Chen** : Conceptualization, Funding acquisition, Supervision, Investigation, Writing – review & editing. **Bo Lu** : Conceptualization, Funding acquisition, Supervision, Investigation, Writing – review & editing.

Funding

This work was supported by National Natural Science Foundation of China (82071239, 82301397), Natural Science Foundation of Ningbo Municipality (2021J309, 2025J060), Ningbo Medical and Health Brand Discipline (PPXK2024-05).

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

No human subjects were involved in this study. All animal experiments were conducted according to a protocol approved by the Institutional Animal Care and Use Committee of Ningbo University (NBU20210351).

Competing interests

The authors declare no competing interests.