T7 peptide-engineered liposomal Irisin mitigates PND progression through AMPK/PGC-1α signaling: multi-omic evidence of metabolic and epigenetic modulation

Abstract

This study explored the molecular mechanisms by which T7 peptide-modified liposomal irisin (T7@Lipo@Irisin) alleviates perioperative neurocognitive disorders (PND) via regulation of the AMPK/PGC-1α metabolic pathway. T7@Lipo@Irisin nanoparticles were prepared by thin-film hydration and ultrasonic dispersion and showed favorable physicochemical performance, with an encapsulation efficiency of approximately 85%. Serum analysis of healthy donors (n = 10) and PND patients (n = 6) showed higher IL-6 and TNF-α and lower brain-derived neurotrophic factor (BDNF) in PND. In vitro, T7@Lipo@Irisin restored mitochondrial membrane potential, reduced reactive oxygen species (ROS) accumulation, enhanced Neuro-2a hippocampal neuron viability, and activated the AMPK/PGC-1α axis under oxidative stress. In a PND mouse model, it improved Garcia neurological scores, preserved neuronal morphology, and decreased apoptosis. Multi-omic integration of scATAC-seq/scRNA-seq and TMT-based proteomics demonstrated enhanced neuro-glial crosstalk, epigenetic activation of metabolic/antioxidant genes (e.g., Sirt1, Nfe2l2), and upregulated pathways (mitochondrial function, NAD-dependent metabolism, synaptic homeostasis). Proteomics confirmed upregulation of SIRT1, NDUFS2, and BDNF, forming a network linked to energy metabolism and neural repair. Collectively, T7@Lipo@Irisin mitigates PND by activating AMPK/PGC-1α to enhance mitochondrial function and stabilize the neuro-microenvironment.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-026-04109-7.

Introduction

Perioperative neurocognitive disorders (PND) encompass a spectrum of surgery-related cognitive impairments, including postoperative delirium, postoperative mild cognitive impairment, and longer-term cognitive decline, and primarily affect middle-aged and elderly adults [1, 2]. PND remains common after general anesthesia in older patients, with an incidence of approximately 20–25% during the first postoperative week, and is associated with delayed recovery, prolonged hospitalization, higher readmission rates, and reduced quality of life [2, 3]. Despite recent advancements in perioperative care, effective preventive and therapeutic strategies for PND remain limited, especially for early postoperative cognitive decline [4]. The pathogenesis of PND is multifaceted, involving neuronal injury, neuroinflammation, mitochondrial dysfunction, and dysregulated neuro-glial interactions, which may progress to irreversible neurodegeneration in severe cases [5, 6]. Given the high burden and the lack of targeted interventions, therapeutic strategies with efficient central nervous system (CNS) delivery and clearly defined mechanisms are needed to support earlier identification and more precise intervention.

Previous studies have identified mitochondrial dysfunction as a key driver of PND, marked by mitochondrial membrane potential loss, reactive oxygen species (ROS) overproduction, reduced adenosine triphosphate (ATP) generation, and pathological opening of the mitochondrial permeability transition pore (mPTP) [7, 8]. These alterations compromise neuronal bioenergetics, intensify oxidative stress, and promote regulated cell death, leading to disruption of neural circuits in cognition-related regions, including the hippocampus and cortex [9]. Moreover, mitochondrial metabolic disturbances impair neuro-glial homeostasis, amplifying inflammatory cytokine release and cytotoxic responses, thereby perpetuating a deleterious feedback loop [10]. Accumulating evidence indicates that re-establishing mitochondrial homeostasis and engaging endogenous antioxidant and bioenergetic programs may mitigate cognitive impairment in PND [11]. No available intervention selectively modulates mitochondrial function within the nervous system, highlighting the need to develop more targeted and effective therapeutic strategies for PND.

AMPK and its downstream transcriptional coactivator PGC-1α form a canonical metabolism-mitochondria regulatory axis that supports energy homeostasis, antioxidant defense, and mitochondrial biogenesis [12]. As an energy sensor, AMPK is rapidly activated through phosphorylation during energy stress and promotes PGC-1α expression, thereby initiating protective metabolic programs that include mitochondrial biogenesis, respiratory chain remodeling, and ROS clearance [13, 14]. In the nervous system, the AMPK/PGC-1α axis has been shown to exert neuroprotective effects in models of cerebral ischemia, Parkinson’s disease, and Alzheimer’s disease [15, 16]. Activation of this pathway alleviates neuronal mitochondrial stress, inhibits apoptosis, and improves behavioral phenotypes, highlighting its substantial therapeutic potential [17, 18]. Clinical translation remains limited by low delivery efficiency, insufficient CNS selectivity, and short pharmacological persistence, which has constrained application in perioperative neuroprotection [19]. Progress in metabolic intervention for PND will require delivery platforms that activate AMPK while enabling efficient transport across the blood-brain barrier (BBB).

Irisin is a hormone-like polypeptide released by skeletal muscle in response to physical activity and has been reported to promote energy-metabolism reprogramming, suppress inflammation, and protect neural function, with these actions closely associated with AMPK/PGC-1α activation [20, 21]. However, Free-Irisin is a small protein with a short in vivo half-life, susceptibility to enzymatic degradation, and limited BBB permeability, which constrains therapeutic utility in CNS disorders [22]. In recent years, nanodelivery systems such as liposomes have been widely employed for CNS drug delivery due to their excellent biocompatibility, tunable properties, and efficient drug-loading capacity [23]. Surface modification with targeting peptides (e.g., transferrin receptor [TfR]-binding T7 peptide) and integration of biomimetic membrane structures (e.g., neuronal membranes) can significantly enhance the brain-targeting and BBB-penetrating capabilities of liposomes, thereby improving their therapeutic efficacy in neurological diseases [24–26]. Mechanistic interrogation can be strengthened by multicellular co-culture Transwell 3D models and single-cell multi-omics approaches, including single-cell RNA sequencing (scRNA-seq) and single-cell Assay for Transposase-Accessible Chromatin using sequencing (scATAC-seq), which enable analysis of neuro-glial interaction dynamics and epigenetic regulatory programs. Taken together, these advances support the development of a biomimetic nanodrug system that improves targeting and controlled release while enabling mechanism-informed evaluation, thereby addressing key barriers that currently limit irisin-based intervention for PND.

Based on the strategy of “brain-targeted liposomal delivery of irisin to activate the AMPK/PGC-1α pathway for PND intervention,” this study constructed T7 peptide-modified liposomes encapsulating irisin (T7@Lipo@Irisin) and evaluated mitochondrial function, oxidative stress, neuroinflammation, and cognitive performance in a PND model. A three-dimensional multicellular co-culture system simulating the BBB was established in vitro to assess the enhanced trans-barrier transport and cell-targeting efficiency conferred by T7 modification and membrane fusion. In vivo, transcriptomic, chromatin accessibility, proteomic, and metabolomic analyses were performed to elucidate how activation of the AMPK/PGC-1α pathway modulates neuro-glial interactions and restores metabolic homeostasis. Additionally, siRNA and CRISPR interference techniques were employed to verify the critical role of this pathway, and the neuroprotective efficacy of irisin-loaded liposomes (Lipo@Irisin) was tested under AMPK-silenced conditions. This study proposed a neuron membrane-mimicking drug delivery platform with precise targeting capability, achieving successful BBB translocation of irisin and mechanistic validation in a PND model. This study proposes a neuron-mimetic drug delivery system with precise targeting capability, demonstrates the feasibility of delivering irisin across the BBB using biomimetic liposomes, and systematically validates the mechanistic basis of irisin in a PND model. These findings further provide theoretical support for AMPK-mediated metabolic intervention in PND. By integrating advances in materials science, neuropharmacological mechanisms, and multi-omics technologies, this strategy offers distinct cross-disciplinary advantages and translational potential and may provide a novel and feasible approach for perioperative neuroprotection.

Materials and methods

Preparation of Lipo@Irisin

Lipo@Irisin was prepared by thin-film hydration followed by ultrasonic dispersion. Briefly, 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), cholesterol, and DSPE-PEG2000-Mal (maleimide-modified phospholipid) were dissolved in chloroform/methanol (2:1, v/v) at a molar ratio of 3:1:0.5. Solvent was removed using a rotary evaporator (Büchi R-300) at 40 °C in a water bath to form a uniform lipid film. The lipid film was then dried under vacuum for at least 2 h to eliminate residual solvent. Hydration was performed with PBS (pH 7.4) containing recombinant human irisin (rh-Irisin; PeproTech; Cat# 130 − 16; 200 µg/mL), followed by incubation with shaking at 37 °C for 30 min to promote lipid dispersion and protein encapsulation. The resulting suspension was ultrasonicated in an ice bath using a Scientz-IID ultrasonic processor (300 W; 2 s on/2 s off; 10 min) to reduce particle size.

After ultrasonication, the suspension was repeatedly extruded (10 cycles) through a 200-nm polycarbonate membrane to obtain uniformly sized irisin-loaded liposomes. All procedures were performed under sterile conditions using EndoFree consumables to minimize endotoxin contamination. Encapsulation efficiency and structural stability of the liposomes were assessed by dynamic light scattering (DLS), transmission electron microscopy (TEM), BCA assay, and high-performance liquid chromatography–tandem mass spectrometry (LC-MS/MS). The preparation was repeated for three independent batches, and batch-to-batch consistency was assessed using the resulting characterization data.

T7 peptide modification and neuronal membrane coating

T7 peptide (HAIYPRH) was conjugated to the liposomal surface through Michael addition between the terminal thiol group of T7 and the maleimide group of DSPE-PEG2000-Mal embedded in the lipid bilayer. The reaction was performed in the dark at 4 °C with gentle shaking for 12 h to maintain peptide stability and reactivity. Unreacted free T7 peptide was removed by ultracentrifugation (20,000 × g, 30 min, 4 °C). Conjugation efficiency was quantified by measuring residual free peptide in the supernatant using the BCA assay. Successful covalent coupling was further supported by Fourier-transform infrared spectroscopy (FTIR), indicated by loss of maleimide-associated peaks and appearance of characteristic peptide-related signals.

For neuronal membrane coating, plasma membranes were isolated from Neuro-2a murine neuroblastoma cells. After PBS washing, cells were lysed in protease inhibitor-containing buffer (50 mM Tris-HCl, pH 7.4; 150 mM NaCl; 1 mM EDTA) using an ultrasonic disrupter on ice (350 W; 2 s on/2 s off; 8 min). The lysate underwent differential centrifugation to enrich plasma membranes, including 1,000 × g to remove nuclei, 10,000 × g to remove mitochondria, and 100,000 × g to collect the membrane fraction. The membrane pellet was dialyzed (MWCO 100 kDa) to reduce nucleic acid and soluble protein contamination, followed by quantification of membrane protein concentration using the BCA assay.

Purified neuronal membranes were mixed with T7-modified liposomes at a 1:1 mass ratio and incubated at 37 °C with shaking for 30 min to facilitate membrane fusion and formation of neuron-mimicking nanocarriers. Fusion and coating were assessed by DLS through changes in particle size and zeta potential, and by TEM to visualize membrane coverage. Western blotting for membrane markers, including Na⁺/K⁺-ATPase and neural cell adhesion molecule (NCAM), was used to confirm incorporation of neuron-derived membrane components on the particle surface. Surface presentation of T7 after membrane fusion was evaluated using a quenching assay with FITC-labeled T7 (FITC-T7) before and after fusion. The modification and coating procedure was repeated in three independent batches to assess the reproducibility of conjugation efficiency and membrane-coating performance.

Particle size, zeta potential, and morphology analysis

Particle size distribution and surface charge were measured by DLS using a Zetasizer Nano ZS90 (Malvern Instruments, UK). Liposome dispersions were diluted in PBS to an appropriate scattering intensity and analyzed at 25 °C to obtain the average hydrodynamic diameter, polydispersity index (PDI), and size distribution. Each sample was measured in triplicate. Zeta potential was determined on the same instrument by electrophoretic light scattering to evaluate surface charge and colloidal stability. Measurements were performed under dust-free conditions, and freshly prepared samples were analyzed within 30 min of dilution.

Morphology was examined by TEM (JEOL JEM-2100, Japan) at 200 kV. Samples were diluted in PBS, deposited onto carbon-coated copper grids for 1 min, and blotted to remove excess liquid. Negative staining was performed using 2% phosphotungstic acid (pH 7.0), followed by air drying. TEM images were used to assess structural integrity, coating uniformity, and particle size, and the observed dimensions were compared with DLS results.

T7 conjugation on the liposome surface was evaluated by FTIR (Nicolet iS50, Thermo Fisher Scientific). Lyophilized samples were mixed with KBr and pressed into pellets. Spectra were collected from 4000 to 400 cm⁻¹, focusing on changes in the maleimide functional peak (~1700 cm⁻¹) and amine-related peptide peaks to verify conjugation efficiency. To further assess drug loading and release behavior of irisin, LC-MS/MS (Agilent 6460 Triple Quadrupole) was used to quantify encapsulation efficiency. Unencapsulated protein was removed by ultrafiltration, and retained fractions were collected for analysis. For release studies, liposomes were incubated in PBS supplemented with 10% fetal bovine serum (FBS) at 37 °C with shaking (100 rpm) for 48 h. Aliquots were collected at predefined time points, and irisin concentration was quantified by LC-MS/MS to generate release profiles and evaluate controlled-release performance. All characterization and release experiments were conducted using three independently prepared batches to assess formulation consistency and process stability.

Evaluation of biomimetic membrane stability and T7 peptide surface exposure

The lipid order and structural stability of the biomimetic membrane were evaluated using Raman spectroscopy. Liposome samples were dropped onto CaF₂ slides and air-dried before analysis with a confocal Raman spectrometer (Renishaw inVia Reflex, UK) using a 785 nm excitation wavelength. Spectra were collected from 400 to 3200 cm⁻¹ using a 10 s integration time and averaged across three consecutive scans. Lipid chain conformation was evaluated by monitoring CH2 stretching near 2850 cm-1 and symmetric CH3 stretching near 2930 cm⁻¹. The intensity ratio (I2850/I2930) was calculated as an index of lipid-order changes associated with membrane packing.

T7 surface exposure after neuronal membrane coating was examined using fluorescence quenching assays with FITC-T7. FITC-T7-modified liposomes were analyzed before and after membrane fusion. Samples were incubated with increasing concentrations of quenchers (benzoate or KI), and fluorescence was measured using a fluorescence spectrophotometer (Hitachi F-7100) at 495 nm excitation and 520 nm emission. Stern-Volmer plots were generated, and Stern-Volmer constants (Ksv) were calculated to compare fluorophore accessibility among groups as a surrogate for T7 exposure on the outer surface. To examine whether membrane coating affected T7-mediated targeting, binding assays were performed using transferrin receptor-positive (TfR⁺) endothelial cells. Cellular binding and distribution were evaluated by confocal microscopy to provide functional evidence of targeting capacity after coating. All assays were conducted in triplicate with independently prepared batches to support reproducibility and statistical reliability.

Peripheral blood sample collection and enzyme-linked immunosorbent assay (ELISA) in clinical PND patients

Tumor samples used in this study were obtained from patients undergoing PND treatment at our hospital between January 2023 and January 2024 (PND group, n = 6), as well as from healthy individuals undergoing routine physical examinations (Donor group, n = 10). Whole blood was collected in serum tubes without anticoagulant or in tubes containing a clot activator. Samples were allowed to clot at 4 °C in an inclined position (45–60°) for 1 h and were then centrifuged at 3,000 rpm for 5 min. The supernatant was collected as serum and stored at 4 °C for short-term use or at −80 °C for long-term storage. Written informed consent was obtained from all participants. The study protocol was approved by the Clinical Ethics Review Committee of our hospital, and followed the principles of the Declaration of Helsinki.

Serum concentrations of IL-6 (Abcam, ab178013), TNF-α (Abcam, ab181421), and BDNF (Abcam, ab212166) were measured using commercial ELISA kits.

Cell culture and experimental grouping

The following cell lines were used: human cerebral microvascular endothelial cells (hCMEC/D3; Cat# SCC066, Merck Millipore), murine microglial cells (BV2; Cat# CL-0493, Guqiao Biotechnology, Shanghai, China), human embryonic kidney cells (HEK 293 T; Cat# CRL-11268, ATCC, USA), and murine neuroblastoma cells (Neuro-2a; Cat# CCL-131, ATCC, USA). All human-derived cell lines were cultured at 37 °C in a humidified atmosphere containing 5% CO₂. hCMEC/D3 cells were maintained in EndoGRO-MV complete medium (customized by Millipore) supplemented with 5% FBS, 1.4 µM hydrocortisone, 5 ng/mL bFGF, 10 mM L-glutamine, 10 µg/mL ascorbic acid, and 1% insulin-transferrin-selenium supplement (ITS-G). BV2 and Neuro-2a cells were cultured in high-glucose DMEM (Gibco, Cat# 11965092) supplemented with 10% FBS. HEK 293 T cells were maintained in antibiotic-free high-glucose DMEM containing 10% FBS and were used for transfection experiments related to liposome construction.

Experimental groups were defined as follows: Control, PBS for 24 h; Free-Irisin, 200 ng/mL irisin added to the upper chamber for 24 h; Lipo@Irisin, 200 ng/mL irisin encapsulated in liposomes for 24 h; T7@Lipo@Irisin, 200 ng/mL irisin encapsulated in T7-modified liposomes (T7 density, 0.5 mol%) for 24 h; siAMPK + T7@Lipo@Irisin, AMPKα siRNA transfection for 24 h followed by treatment with T7@Lipo@Irisin (200 ng/mL irisin; T7 density, 0.5 mol%) for an additional 24 h.

Construction of multicellular co-culture transwell model

A three-dimensional multicellular co-culture BBB model was established using a Transwell system. Transwell inserts (Corning Costar) with polycarbonate membranes (0.4 μm pore size; 12 mm diameter) served as the scaffold. hCMEC/D3 cells were seeded in the upper (apical) chamber at 3 × 10⁴ cells/cm². The membrane surface was pre-coated with extracellular matrix (ECM) gel to enhance cell adhesion and promote monolayer formation. Simultaneously, primary murine cortical neurons and BV2 microglia were co-seeded into the lower (basal) chamber at a 1:1 ratio (total 1 × 10⁵ cells/well) using neuron-specific medium (Neurobasal supplemented with B27 and GlutaMAX). All cell lines were obtained from the institutional cell repository, authenticated by STR profiling, and confirmed to be mycoplasma-free. Each experiment was performed in triplicate.

After seeding, the co-culture model was maintained under standard conditions (37 °C, 5% CO₂) for 7 consecutive days, with medium replacement in both chambers every 48 h. Barrier integrity was assessed by transepithelial electrical resistance (TEER) using a Millicell ERS-2 V-ohmmeter (Merck Millipore). A TEER value exceeding 250 Ω·cm² was used as the criterion for successful barrier formation. Endothelial tight-junction formation was further evaluated by immunofluorescence staining for ZO-1 and Occludin. In addition, permeability was assessed using FITC-dextran (40 kDa; final concentration 1 mg/mL) to quantify paracellular flux from the apical to the basal chamber.

Analysis of liposome uptake and targeting specificity

Cellular uptake and targeting specificity of liposomes in the BBB co-culture model were assessed using DiI-labeled formulations (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate; Thermo Fisher Scientific). After labeling, unbound dye was removed by ultrafiltration, and labeled liposomes were resuspended in serum-free medium. DiI-labeled liposomes were added to the apical chamber at a final concentration of 200 µg/mL and incubated for 4 h. Cells were then fixed with 4% paraformaldehyde. hCMEC/D3 cells from the apical chamber and neurons from the basal chamber were collected for imaging and quantitative analyses.

Uptake behavior was visualized using a laser scanning confocal microscope (Leica TCS SP8). Immunofluorescence co-staining was performed with TfR to identify endothelial cells and NeuN as a neuronal nuclear marker. Nuclei were counterstained with DAPI. Acquisition parameters, including excitation power and exposure settings, were kept constant across groups. Semi-quantitative analysis was performed in ImageJ (NIH) by measuring red fluorescence intensity per unit area to compare uptake across treatments. Flow cytometry (BD LSRFortessa) was used for quantitative assessment. Single-cell suspensions of endothelial cells and neurons were prepared, and dead cells were excluded before analysis. The percentage of DiI-positive cells and mean fluorescence intensity (MFI) were recorded to evaluate whether T7 modification enhanced cell-targeting specificity. Experiments were performed in triplicate. Data were reported as mean ± standard deviation, and two-way ANOVA was used to test interaction effects between liposome formulation and cell type. A two-sided p-value < 0.05 was considered statistically significant.

Assessment of BBB integrity

BBB integrity after liposome exposure was evaluated using complementary electrophysiological, tracer-permeability, and molecular assays. TEER was measured with a Millicell ERS-2 voltohmmeter (Merck Millipore) to assess tight-junction integrity in the Transwell endothelial monolayer. TEER was recorded at baseline (0 h) and at 4 h and 8 h after liposome treatment. Measurements were performed at 37 °C, and background resistance from blank inserts was subtracted. TEER values were reported as Ω·cm², with each group consisting of at least three independent replicates. A TEER reduction greater than 20% was used as the criterion for substantial barrier disruption.

Paracellular permeability was assessed using 40 kDa FITC-dextran (Sigma-Aldrich, F8019). FITC-dextran was added to the apical chamber at 1 mg/mL. After 4 h, medium from the basal chamber was collected, and fluorescence was measured using a microplate reader (excitation/emission, 490/520 nm). A standard curve was used to calculate permeability per unit area (µg·cm⁻²·h⁻¹).

Tight-junction protein expression was evaluated by Western blot. hCMEC/D3 cells were harvested from the Transwell membrane, and total protein was extracted. ZO-1 (Thermo Fisher, 61–7300) and Occludin (Abcam, ab167161) were detected using specific primary antibodies, and GAPDH served as the loading control. TEER, FITC-dextran permeability, and Western blot results were analyzed together to support a multi-level assessment of barrier integrity. Experiments were repeated in three independently cultured batches. Data were tested for normality and analyzed by one-way ANOVA to compare groups.

Assessment of in vitro neuroprotective effects

Neuroprotective effects of Lipo@Irisin were evaluated in primary murine neurons isolated from the co-culture system by measuring mitochondrial function and cell-homeostasis endpoints. Mitochondrial membrane potential was assessed using JC-1 (Beyotime, C2006). After PBS washing, neurons were incubated with JC-1 working solution (2 µM) at 37 °C for 20 min. Fluorescence was measured using a microplate reader, with excitation/emission at 488/530 nm for JC-1 monomers and 525/590 nm for JC-1 aggregates. The aggregate-to-monomer (red-to-green) fluorescence ratio was calculated, and a lower ratio indicated mitochondrial depolarization.

Mitochondrial ROS levels were measured using MitoSOX Red (Thermo Fisher, M36008). Neurons were incubated with MitoSOX (5 µM) for 10 min, washed with PBS, and analyzed by flow cytometry (BD LSRFortessa). Red fluorescence was detected at 510/580 nm (Ex/Em) to reflect mitochondrial oxidative stress. Intracellular ATP levels were quantified using the Luminescent ATP Assay Kit (Promega, Cat# G7570), which is based on the luciferase-luciferin bioluminescence reaction. Cell lysates were mixed with detection reagent, and luminescence was recorded immediately. ATP concentrations were calculated from a standard curve and expressed as nmol/10⁶ cells.

For mechanistic analysis, total protein was extracted for Western blotting. Targets included AMPK/PGC-1α pathway components (p-AMPK, AMPKα, and PGC-1α) and apoptosis-related proteins (Bcl-2 and Bax), with GAPDH as the loading control. All experiments were repeated in triplicate. Results were expressed as mean ± standard deviation, and statistical differences were assessed using one-way ANOVA or two-way ANOVA when multiple factors were involved. A p-value < 0.05 was considered statistically significant.

SiRNA and CRISPR-mediated interference of AMPK/PGC-1α expression

To determine whether liposome-mediated neuroprotection required AMPK/PGC-1α signaling, AMPKα and PGC-1α were suppressed in hCMEC/D3 cells and primary neurons using siRNA and CRISPR/Cas9 approaches. For transient knockdown, gene-specific siRNAs targeting human AMPKα (PRKAA1) and PGC-1α (PPARGC1A) were synthesized by GenePharma (purity > 90%) and transfected into cells at a final concentration of 50 nM. Transfection was performed in cells at logarithmic growth phase using Lipofectamine 3000 (Thermo Fisher Scientific, L3000015). After 6 h incubation in antibiotic-free medium, fresh complete medium was added. Transfection performance and cytotoxicity were assessed at 24 h post-transfection. A negative control siRNA group (si-NC) and an untreated control group were included in each experiment.

For stable knockout, a CRISPR/Cas9 system was designed to target critical exonic regions of AMPKα and PGC-1α. Guide RNAs (gRNAs) were cloned into the lentiCRISPR v2 vector (Addgene #52961), packaged into lentiviral particles using HEK 293 T cells, and used to infect target cells. Following infection, cells were selected with 1 µg/mL puromycin for 7 days to establish stable knockout lines. Gene knockdown or knockout efficiency was first assessed at the mRNA level using qRT-PCR (SYBR Green system, Thermo Fisher). Sangon Biotech synthesized primers, and relative gene expression was calculated using the 2^-ΔΔCt method. Protein-level validation was performed by Western blot using antibodies against AMPKα, p-AMPK, and PGC-1α, with GAPDH as the internal loading control.

After pathway interference, neuroprotective readouts for Lipo@Irisin were re-evaluated, including mitochondrial membrane potential, ATP production, ROS accumulation, and apoptosis-related protein expression, to determine pathway dependence. Transfection and gene-editing experiments were independently repeated at least three times, and group comparisons were analyzed by one-way ANOVA.

Assessment of pathway activation levels

To systematically evaluate the activation of the AMPK/PGC-1α signaling pathway following liposome-mediated irisin delivery, both the expression levels and subcellular localization of key signaling proteins were examined. After treatment, total cellular proteins were extracted using RIPA lysis buffer and quantified using the BCA assay. Western blot analysis was performed to detect levels of p-AMPK (Thr172), total AMPKα, PGC-1α, UCP2, SIRT1, and Nrf2.

Nrf2 subcellular localization was examined by separating cytoplasmic and nuclear fractions using a Nuclear/Cytoplasmic Extraction Kit (Thermo Fisher Scientific, Cat# 78833). Fraction purity was verified by Lamin A/C as a nuclear marker and GAPDH as a cytoplasmic marker. Nrf2 levels in nuclear and cytoplasmic fractions were quantified to determine whether liposome treatment promoted nuclear translocation associated with transcriptional activation.

In addition, the expression of UCP2 and SIRT1 was measured to assess mitochondrial pathway responses, providing insight into whether irisin exerts mitochondrial protection through modulation of energy metabolism-related regulators. All samples were derived from three independent experimental replicates, and data were statistically analyzed using GraphPad Prism.

Mitochondrial metabolic analysis

To evaluate the impact of Lipo@Irisin on cellular energy metabolism reprogramming, mitochondrial oxidative phosphorylation and glycolytic function in neurons were monitored in real-time using a Seahorse XF extracellular flux analyzer (Agilent Technologies XF96). Cells were seeded in Seahorse XF96 microplates at a density of 2 × 10⁴ cells per well and cultured until stable adherence. Before analysis, the growth medium was replaced with XF Base Medium supplemented with 10 mM glucose, 1 mM pyruvate, and 2 mM glutamine. Plates were equilibrated for 1 h at 37 °C in a non-CO₂ incubator. Oxygen consumption rate (OCR) was measured using the Mito Stress Test with sequential injections of oligomycin (1 µM), FCCP (1.5 µM), and antimycin A/rotenone (0.5 µM). Basal respiration, maximal respiration, ATP production-linked respiration, and non-mitochondrial respiration were calculated from OCR profiles. Glycolytic capacity and stress responsiveness were assessed by extracellular acidification rate (ECAR) using the Cell Energy Phenotype Test module, which recorded ECAR at baseline and after metabolic stress to quantify shifts between oxidative phosphorylation and glycolysis.

Targeted metabolomics was performed to assess tricarboxylic acid (TCA) cycle activity after liposomal treatment. Cellular metabolites were extracted in ice-cold methanol/water (80:20), clarified by centrifugation, dried, and reconstituted in 50% methanol. Quantification was carried out on an Agilent 1290 UHPLC system coupled to an Agilent 6460 triple quadrupole mass spectrometer to measure citrate, isocitrate, succinate, malate, oxaloacetate, and α-ketoglutarate. Stable isotope-labeled internal standards were included to improve quantification accuracy. Metabolite abundance was calculated from peak areas in MassHunter software using external standard curves and normalized to protein content, with results reported as nmol/mg protein. Seahorse and LC-MS/MS results were interpreted together to evaluate changes in mitochondrial metabolism and overall energy-metabolism reprogramming after Lipo@Irisin treatment. Experiments were independently repeated three times, and statistical comparisons were performed by ANOVA with a two-sided p-value < 0.05 as the significance threshold.

Brain tissue preparation and cell isolation

The PND mouse model was established using 8–10-week-old male C57BL/6J mice. Cortical and hippocampal tissues were harvested on postoperative day 7 from control and PND groups. Each group included three biological replicates at the collection time point. Brain regions from each mouse were processed independently, and tissue pooling was not performed. After deep anesthesia, mice underwent intracardiac perfusion with pre-chilled HBSS. Cortex and hippocampus were rapidly dissected and kept on ice.

Dissected tissues were minced and enzymatically digested at 37 °C for 30 min in collagenase IV (1 mg/mL) and DNase I (20 U/mL), with gentle trituration every 10 min. The suspension was filtered through a 70 μm strainer to generate a single-cell suspension. Myelin and debris were removed by 30%/70% Percoll density-gradient centrifugation. The interface cell layer was collected, washed twice with HBSS, and resuspended in PBS containing 0.04% BSA. Cell viability was assessed by Trypan Blue staining, and viability exceeded 85% in all samples.

Cell nuclei were prepared by gentle lysis, filtered through a 40 µm strainer, and counted after DAPI staining. Only samples with a nuclear viability above 85% were included. Both single-cell and nuclear suspensions were adjusted to approximately 1000 cells or nuclei per µL and loaded onto the 10x Genomics Chromium Controller. scRNA-seq libraries were prepared using the Chromium Single Cell 3’ v3.1 Reagent Kit, whereas scATAC-seq libraries were generated with the Chromium Single Cell ATAC Kit.

To minimize the loss of individual variability, samples were multiplexed using CellPlex/antibody hashing (Hashtag-oligo) prior to loading, and computational demultiplexing was applied to assign sample origin. All libraries were sequenced on the Illumina NovaSeq 6000 platform. For scRNA-seq, sequencing was performed in PE28 × 91 mode with a target depth of ≥ 30,000 reads per cell, while scATAC-seq used PE50 × 50 mode with a target of ≥ 35,000 fragments per nucleus. PhiX was included as an internal control.

Quality control and single-cell data analysis

Raw sequencing data were aligned to the mouse reference genome (mm10) using Cell Ranger (scRNA v6.x; scATAC v2.x). Low-quality cells with either abnormally low or high gene counts or a mitochondrial fraction > 10% were excluded. Only high-quality nuclei with TSS enrichment ≥ 8, fragments-in-peaks ≥ 3000, and nucleosome signal ≤ 4 were retained. Doublets were predicted and removed using Scrublet/DoubletFinder. For scATAC-seq, MACS2 was used to generate a unified peak set across samples, followed by construction of a peak × cell matrix, TF-IDF normalization, and latent semantic indexing (LSI) for dimensionality reduction. Harmony was applied for batch-effect correction. Clustering results were visualized using UMAP, and cell-type annotation was performed with reference to a published hippocampal single-cell atlas [27]. Major populations included neurons, astrocytes, oligodendrocytes, microglia, and endothelial cells. Within each annotated cell type, differential chromatin accessibility between control and PND groups was tested using the Wilcoxon rank-sum test with Benjamini-Hochberg correction (q < 0.05). Genes mapped to differential peaks were subjected to Gene Ontology (GO)/Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis to identify programs related to mitochondrial function, ROS clearance, and synaptic homeostasis.

ATAC-seq key loci and interaction analysis

Nuclei from hippocampal and cortical tissues in control and PND groups were processed for scATAC-seq using the 10x Genomics Chromium platform (target ≥ 25-50k fragments per cell). Alignment and quality control were performed with Cell Ranger ATAC, and doublets were filtered using ArchR scores. After unified peak set construction, dimensionality reduction, and batch correction with LSI/Harmony, annotation was performed. CoveragePlot/IGV was used to visualize intergroup differences in accessibility across ±50–100 kb windows surrounding the Sirt1 and Nfe2l2 (Nrf2) loci, with reference to ENCODE annotations. Cicero co-accessibility analysis (make_cicero_cds/run_cicero) was applied, defining high-confidence enhancer-promoter connections as those with a co-accessibility score > 0.35. Genes associated with “new or strengthened interactions” were subjected to GO/KEGG enrichment (q < 0.05). Motif enrichment was assessed with HOMER/FindMotifs, and transcription factor activity changes were evaluated by chromVAR deviation scores, highlighting NRF1, CREB1, and SP1 as key regulators.

Proteomic differential analysis

To support transcriptomic and epigenomic observations, TMT-labeled quantitative proteomics was performed on matched cortical and hippocampal tissue samples from the same mouse batch. Tissues were homogenized, lysed, and total protein was quantified. Proteins were digested with trypsin, and peptides were labeled using a TMT 10-plex kit (Thermo Fisher Scientific). Labeled peptides were pooled, fractionated by high-performance liquid chromatography (HPLC), and analyzed by high-resolution tandem mass spectrometry on an Orbitrap Fusion Lumos instrument. Raw spectra were processed in Proteome Discoverer (v2.5, Thermo Fisher Scientific) for peptide identification and database searching against the UniProt mouse reference proteome. Quantitative data were normalized and subjected to differential expression analysis using limma (v3.46.0, R/Bioconductor), with thresholds set at |log₂FC| > 0.58 and p < 0.05. Volcano plots were generated to visualize group differences, and proteins of interest, including SIRT1, NDUFS2, and BDNF, were highlighted to support multi-omics integration with the single-cell datasets.

STRING protein interaction network analysis

Significantly upregulated proteins were input into the STRING database (version 12.0, https://string-db.org/) with the species set to Mus musculus. The confidence threshold was set to 0.7 (high confidence), and the interaction network was exported. Network visualization and module partitioning were conducted using STRING’s built-in tools, and enriched pathways were annotated through its GO/KEGG analysis functions.

Establishment of the PND model and animal grouping

Eight-week-old male C57BL/6J mice (22–26 g) were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. All procedures were approved by the Institutional Animal Care and Use Committee. After 7 days of acclimatization, mice were randomized into experimental groups. For each group, 12 mice were allocated for behavioral testing and an additional 6 mice were allocated for other analyses. All surgical procedures were performed under sterile conditions. After intraperitoneal anesthesia with sodium pentobarbital (50 mg/kg), mice were placed on a surgical board. A PND model was induced by small intestinal resection. A 1.5 cm midline abdominal incision was made, approximately 30% of the jejunum was gently exteriorized and resected, and end-to-end anastomosis was completed before returning the intestine to the abdominal cavity. Postoperative fluid resuscitation was performed with 1 mL prewarmed saline. Sham mice underwent laparotomy without intestinal resection. After surgery, mice were transferred to a temperature-controlled recovery chamber for monitoring and then received assigned interventions at predefined time points. Six groups were included: Sham (NS), PND (surgery + NS), PND + Free-Irisin (surgery + Free-Irisin), PND + Lipo@Irisin (surgery + Lipo@Irisin), PND + T7@Lipo@Irisin (surgery + T7@Lipo@Irisin), and PND + T7@Lipo@Irisin + siAMPK (siAMPK + surgery + T7@Lipo@Irisin). Free-Irisin, Lipo@Irisin, and T7@Lipo@Irisin were administered by tail vein injection on postoperative days 1, 3, and 7. Liposomal formulations were administered at 200 µg/kg (Irisin-equivalent), and Free-Irisin was administered at 3 µg/g. siRNA intervention was conducted using AMPKα siRNA delivered via Entranster-in vivo transfection reagent through a single intracerebroventricular injection 3 days before surgery. TFR1 antibody was administered at 6 mg/kg through tail vein injection 12 h prior to T7@Lipo@Irisin administration. On postoperative day 7, brain tissue, blood samples, and behavioral data were collected for pathological, functional, and multi-omics analyses. All samples were collected in the same batch to avoid batch-effect interference.

Drug delivery and imaging analysis

To assess brain-targeting distribution and systemic safety in vivo, Lipo@Irisin and T7@Lipo@Irisin were labeled with the near-infrared dye DiR (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindotricarbocyanine iodide; Thermo Fisher Scientific). DiR was added to the organic lipid phase at 0.5% (w/w) and co-evaporated during thin-film formation. After liposome preparation, unincorporated dye was removed by ultrafiltration centrifugation. DiR-labeled liposomes were administered by tail vein injection at 1 µg/g body weight in a total volume of 200 µL. Whole-body fluorescence imaging was performed at 1 h, 6 h, and 24 h after injection using an IVIS Spectrum in vivo imaging system (PerkinElmer) with excitation at 748 nm and emission at 780 nm. Mice were anesthetized and imaged in a standardized position. Fluorescence signals were quantified in Living Image software, with region-of-interest analysis applied to the brain and major organs, including liver, spleen, kidneys, and lungs, to compare central distribution between formulations.

Systemic toxicity was evaluated in the same animals after imaging. MicroCT scanning was performed using a Bruker SkyScan 1176 system to assess gross organ volume and structural integrity of the liver, kidneys, and lungs. Blood was then collected for serum biochemistry, including alanine aminotransferase (ALT) and aspartate aminotransferase (AST) as liver injury indicators and blood urea nitrogen (BUN) and creatinine (Cr) as renal function indicators, measured on an automated biochemical analyzer (Beckman AU5800). Major organs, including the liver, kidney, heart, and lung, were harvested for hematoxylin and eosin (H&E) staining to evaluate histopathological changes. All data were derived from three independent experiments, with each group containing n = 6 mice for imaging and toxicity assessments. Fluorescence intensity and biochemical data were expressed as mean ± standard deviation. Intergroup comparisons were conducted using one-way ANOVA, with p < 0.05 considered statistically significant. These experiments evaluated whether T7 modification increased brain-targeting efficiency while maintaining acceptable biocompatibility and low systemic toxicity.

Histopathology and immunohistochemistry (IHC) of brain tissue

To evaluate brain pathology and neuronal injury after liposomal intervention in the PND model, hippocampal and cortical tissues were collected on postoperative day 7. Samples were paraffin-embedded and serially sectioned at 5 μm. Sections were deparaffinized in xylene and rehydrated through graded ethanol. H&E staining and Nissl staining with 0.1% toluidine blue were performed. H&E staining was used to evaluate overall tissue architecture, neuronal arrangement, and edema. Nissl staining was used to assess neuronal soma morphology and Nissl body distribution. For each section, three random high-power fields (×400) were selected to quantify degenerative features, including neuronal disorganization, vacuolar degeneration, and cytoplasmic pallor, as structural indices of neuronal injury.

IHC was performed to assess markers related to neuronal survival, glial activation, and programmed cell death. Sections were incubated overnight at 4 °C with primary antibodies against NeuN, GFAP, Iba1, and cleaved caspase-3. After washing, HRP-conjugated secondary antibodies were applied, followed by DAB development and hematoxylin counterstaining before mounting. Antibodies were obtained from CST or Abcam, and working dilutions were optimized based on datasheets and preliminary validation. Images were acquired using a Nikon Eclipse 80i microscope. Quantification was performed in ImageJ by calculating the positive staining area (% area) and mean integrated optical density (IOD/area). For each mouse, values were averaged across three fields from multiple sections. Immunostaining was performed using independent biological samples in triplicate to support reproducibility. Group comparisons were analyzed by one-way ANOVA with Tukey’s post hoc test, with p < 0.05 considered statistically significant.

Assessment of mitochondrial function

Hippocampal and cortical tissues were collected on postoperative day 7. Mitochondria were isolated using a mitochondrial isolation kit (Beyotime, C3606). Tissues were homogenized in pre-chilled lysis buffer and processed at 4 °C according to the protocol provided by the manufacturer. Homogenates were centrifuged at 600 × g for 10 min to remove nuclei and debris, and supernatants were centrifuged at 11,000 × g for 10 min to pellet mitochondria. Mitochondrial pellets were resuspended, and protein concentrations were determined using a BCA assay for normalization of downstream assays.

Mitochondrial membrane potential was assessed using the JC-1 probe (Beyotime, C2006). Equal amounts of mitochondrial protein were resuspended in JC-1 working solution and incubated at 37 °C for 20 min. Red/green fluorescence ratios were measured with a microplate reader (Ex/Em = 488/530 nm for JC-1 monomers and 525/590 nm for aggregates); a decreased ratio indicated mitochondrial depolarization. ATP levels were quantified using a luciferase-based assay (Promega, Cat# G7570), and luminescence intensity was used to reflect energy metabolism status. ATP concentration was expressed as nmol/mg protein. ROS levels were measured using the MitoSOX Red fluorescent probe (Thermo Fisher, M36008). Mitochondria were stained at a final concentration of 5 µM for 10 min, and fluorescence intensity was recorded at Ex/Em = 510/580 nm.

mPTP opening was evaluated using a Ca²⁺-induced OD540 absorbance assay (Beyotime, C2009). Equal amounts of mitochondrial protein (10–15 µg) were added to reaction buffer containing 250 mM sucrose, 10 mM MOPS, 5 mM succinate, and 1 mM phosphate, followed by addition of CaCl2 (final 200 µM) to initiate pore opening. Absorbance at 540 nm was measured using a microplate reader (SpectraMax iD3) at 0, 2, 4, 6, 8, 10, and 12 min. A faster decline in OD540 reflected greater mPTP opening. Analyses included Sham, PND, PND + Lipo@Irisin, and PND + T7@Lipo@Irisin groups. Negative controls without Ca²⁺ and positive controls with FCCP or H2O2 were included to verify assay performance. Time-OD curves were generated, and the area under the curve (AUC) was calculated to compare mitochondrial membrane stability among groups. All experiments were independently repeated three times.

Western blot and RT-qPCR analysis

To clarify the effects of liposomal intervention on neuroprotective signaling pathways and apoptosis-regulating factors, both protein and mRNA expression levels of the AMPK/PGC-1α signaling axis and its downstream targets were analyzed. For Western blotting, total protein was extracted from brain tissues or cultured neurons using RIPA lysis buffer, and protein concentrations were determined by BCA assay. Equal amounts of protein were separated on 10% SDS-PAGE gels and transferred to PVDF membranes. Membranes were blocked in 5% non-fat milk for 1 h and incubated overnight at 4 °C with primary antibodies. After washing, membranes were incubated with HRP-conjugated secondary antibodies (1:5,000) and developed using an ECL chemiluminescence detection system. Band intensities were quantified in ImageJ and normalized to GAPDH. Antibody information is provided in Table S1.

For RT-qPCR, total RNA was extracted by TRIzol and assessed for purity and integrity using a NanoDrop 2000 spectrophotometer and 1% agarose gel electrophoresis. First-strand cDNA was synthesized using HiScript III RT SuperMix (Vazyme). Quantitative PCR was performed with a SYBR Green system (Applied Biosystems) in a 10 µL reaction volume. Primers were designed and synthesized by Sangon Biotech and were validated for specificity in preliminary assays. Cycling conditions included 95 °C for 30 s, followed by 40 cycles of 95 °C for 10 s and 60 °C for 30 s. Each sample was analyzed in triplicate as technical replicates. Relative expression was calculated using the 2^-ΔΔCt method with GAPDH as the reference gene. Primer sequences are listed in Table S2. Protein and mRNA results were interpreted together to support multi-level consistency in pathway-related changes.

To minimize batch effects, the same reference sample was included in each Western blot and RT-qPCR run. Runs were included for analysis only when the reference sample showed stable performance across experiments.

Morris water maze (MWM) test

Spatial learning and memory were evaluated using the MWM starting on postoperative day 7. The apparatus consisted of a circular pool (120 cm diameter, 50 cm depth) filled with water maintained at 22 ± 1 °C. Latex paint was added to render the water opaque. A circular platform (10 cm diameter) was placed in a fixed quadrant and submerged 1 cm below the water surface. Room lighting, distal visual cues, and the testing environment were kept constant throughout the experiment to maintain stable spatial references.

Training was conducted over 5 consecutive days. Each mouse completed four trials per day, with different start locations and 15 min inter-trial intervals. Each trial lasted up to 60 s. Escape latency was recorded when the mouse reached the platform, followed by a 10 s stay on the platform. When a mouse did not locate the platform within 60 s, the mouse was guided to the platform and allowed to remain for 30 s. Swim trajectories, velocity, and search patterns were recorded using an automated video tracking system (ANY-maze, Stoelting Co.). A probe trial was performed on day 6 with the platform removed. Mice were released from the quadrant opposite the target quadrant. Outcome measures included platform-site crossings, latency to first enter the former platform zone, and the percentage of time spent in the target quadrant. Behavioral testing and analysis were performed under blinded conditions. Sample size for behavioral tests was determined based on statistical power analysis, with 12 mice included in each group. Group differences were analyzed using one-way ANOVA or the Kruskal–Wallis test, followed by Benjamini–Hochberg (BH) correction. A q-value < 0.05 was considered statistically significant.

Y-maze and novel object recognition (NOR) tests

Short-term working memory and recognition memory were evaluated using the Y-maze spontaneous alternation test and the NOR test. The Y-maze consisted of three arms (A, B, and C), each 40 cm in length, arranged at 120° angles. Mice underwent a 5 min habituation session on the day before testing. For the test, each mouse was placed at the end of arm A and allowed to freely explore for 8 min. Spontaneous alternation was defined as successive entries into three different arms (A-B-C or C-A-B). The alternation rate was calculated as (number of alternations)/(total arm entries −2) × 100%. Behavior was recorded and analyzed using an automated tracking system (ANY-maze), and experimenters were blinded to group assignments.

The NOR test was performed in the same behavioral chamber and included a familiarization phase and a test phase. During familiarization, each mouse explored two identical objects (A1 and A2) for 5 min. After a 1 h retention interval, one familiar object was replaced by a novel object (B) that differed in shape, color, or texture, and exploration time for each object was recorded during a 5 min test. Recognition memory was quantified using the discrimination index (DI): (time with novel object − time with familiar object)/(total exploration time) × 100%. Higher DI values reflect stronger preference for the novel object and better recognition memory.

Each group included 12 mice based on power calculations. Mice with abnormal locomotor activity during acclimation or with total exploration time below 10 s were excluded from analysis. Tests were performed in a quiet, temperature-controlled environment. Behavioral outcomes are presented as mean ± standard deviation. Group differences were assessed using one-way ANOVA or the nonparametric Kruskal–Wallis test, as appropriate. Multiple comparisons were corrected using the BH procedure, and a q-value < 0.05 was considered statistically significant.

Garcia scoring system

Neurological function in PND mice was assessed on postoperative day 7 using the Garcia neurological scoring system. Six domains were evaluated: (1) spontaneous activity (0–3), (2) contralateral forelimb grasp strength (0–3), (3) contralateral tactile response (0–3), (4) limb extension reflex (0–3), (5) gait and walking coordination (0–3), and (6) beam balance performance (0–3). Scores across domains were summed to generate a total score ranging from 3 to 18, with higher scores indicating better neurological function.

All assessments were performed by two trained investigators who were blinded to group allocation. The final score for each mouse was calculated as the mean of the two raters to reduce subjective bias. Mice were allowed to acclimate for 5 min before testing, and domains were evaluated in a fixed sequence. Total testing time was limited to 15 min per mouse. Sessions were video-recorded, and blinded review was used for score confirmation when needed. A total of 12 animals were included in each group. Scoring data are presented as mean ± standard deviation. Functional recovery over time was evaluated using two-way repeated-measures ANOVA (Two-way RM ANOVA), and p-values were adjusted using the BH procedure. A q-value < 0.05 was considered statistically significant.

Evaluation of intervention effects under AMPK Silencing

To determine whether the neuroprotective effects of T7@Lipo@Irisin are dependent on AMPK signaling, an siRNA-mediated AMPK-silenced model was established, and the liposomal intervention was repeated under these conditions. Male C57BL/6J mice (8 weeks old) received intracerebroventricular delivery of AMPKα-specific siRNA (10 µg/5 µL; RiboBio) 3 days before surgery using Entranster-in vivo transfection reagent. Injections were performed under stereotaxic guidance with coordinates based on the Paxinos atlas (AP −0.5 mm, ML ±1.0 mm, DV −2.5 mm). Mice were monitored for 72 h after injection to confirm normal behavior and absence of acute neurological deficits. On postoperative day 1, AMPK knockdown in cortical and hippocampal tissues was verified by RT-qPCR and Western blotting. Only mice with knockdown efficiency greater than 80% were included in subsequent experiments.

Under AMPK-silenced conditions, T7@Lipo@Irisin was administered via tail vein injection (1 µg/g body weight) on the day of surgery. Comparative groups included a non-intervention PND model, a liposome-treated PND group, and a Sham-operated group. Intervention outcomes were evaluated across multiple domains: neuronal pathology (Nissl staining and cleaved caspase-3 IHC), mitochondrial function (JC-1 staining, ATP levels, mPTP opening), protein expression along the AMPK/PGC-1α axis, and behavioral performance (MWM and Garcia scoring). Each group consisted of six mice, and all experiments were repeated across three independent biological replicates. Behavioral assessments were conducted under blinded conditions. Data were expressed as mean ± standard deviation. Statistical analysis was performed using one-way ANOVA followed by Tukey’s post hoc test.

Evaluation of systemic toxicity and biosafety

Systemic toxicity and biosafety of T7@Lipo@Irisin were evaluated on postoperative days 7 and 14 using serum biochemistry and histopathological assessment. Peripheral blood and major organs were collected from mice at each time point. Blood was obtained by retro-orbital sampling and centrifuged at 3,500 × g for 10 min to isolate serum. ALT and AST were measured as indices of liver function, and BUN and Cr were measured as indices of renal function using an automated biochemical analyzer (Beckman AU480) with colorimetric assays. Serum complement C3a was quantified by ELISA (Thermo, EEL091). Sample handling followed low-temperature and light-protected conditions to maintain analyte stability.

Following PBS perfusion, liver, kidney, heart, and lung tissues were harvested, paraffin-embedded, sectioned at 4 μm thickness, and subjected to H&E staining. Histological evaluation under light microscopy focused on cellular integrity, nuclear structure, edema, hemorrhage, necrosis, and inflammatory cell infiltration to assess systemic tissue toxicity. For each organ, three non-overlapping fields per section were scored. Pathological injury was quantified semi-quantitatively on a scale from 0 to 4, and two blinded pathologists independently evaluated all slides.

Each group consisted of six mice. Data were expressed as mean ± standard deviation. Statistical significance was determined using one-way ANOVA followed by Bonferroni post hoc testing, with p < 0.05 considered significant.

Results

Successful construction of T7@Lipo@Irisin and membrane fusion

Lipo@Irisin was prepared by thin-film hydration with ultrasonic dispersion, followed by T7 peptide conjugation and neuronal membrane coating to generate the neuron-mimetic nanocarrier T7@Lipo@Irisin with stable physicochemical properties and uniform morphology (Fig. 1 A). DLS analysis revealed an average particle size of 168.2 ± 5.6 nm, a PDI of 0.168, and a zeta potential of −23.7 ± 1.2 mV, indicating good colloidal stability under physiological conditions (Fig. 1B). TEM further confirmed a spherical morphology with a clear outer coating and a bilayer-like membrane structure (Fig. 1 C). FTIR supported successful T7 conjugation. The maleimide-associated peak at 1685 cm⁻¹ detected in unmodified liposomes was not observed after conjugation, and peptide-related amide I and II signals became more evident (Fig. 1D). Conjugation efficiency, calculated from the reduction of free T7 peptide measured by BCA assay before and after reaction, reached 78.6%, with intra-batch variation below 5% across three independent preparations (Fig. 1E). Membrane fusion efficiency analysis showed a slight increase in particle size to 178.4 ± 4.9 nm after neuronal membrane encapsulation, with minimal change in zeta potential. Western blot analysis revealed significant expression of membrane-specific proteins Na⁺/K⁺-ATPase and NCAM, indicating good retention of membrane-origin characteristics (Fig. 1 F).

Fig. 1.

Construction and Characterization of Lipo@Irisin. Note: (A) Schematic diagram illustrating the construction of T7@Lipo@Irisin, including key steps such as lipid membrane formation, T7 peptide conjugation, and neuron membrane fusion; (B) Particle size distribution and zeta potential measured by DLS, showing a uniform nanoparticle size after formulation; (C) TEM image displaying the morphology of liposomes with a characteristic bilayered spherical structure; (D) FTIR spectra indicating the disappearance of the maleimide peak, confirming successful T7 conjugation; (E) T7 conjugation efficiency assessed using the BCA method, demonstrating good batch-to-batch reproducibility; (F) Western blot analysis comparing the expression of membrane proteins such as Na⁺/K⁺-ATPase and NCAM before and after membrane fusion, confirming the integrity of neuron membrane incorporation

Liposomes exhibit high encapsulation efficiency and Sustained-Release properties for Irisin

Encapsulation efficiency and in vitro release behavior of the liposomal irisin system were evaluated using LC-MS/MS quantification, dialysis-based release under simulated physiological conditions, kinetic fitting, and particle-size stability testing (Fig. 2 A). LC-MS/MS showed an encapsulation efficiency of 84.1 ± 2.3% for T7@Lipo@Irisin, supporting formulation robustness across independent preparations (Fig. 2B). Further assessment of sustained-release performance was conducted using the dialysis bag method under simulated physiological conditions (PBS +10% FBS, 37 °C). Free-Irisin showed rapid release, reaching 78.3% at 24 h and 89.4% at 48 h. In contrast, T7@Lipo@Irisin released irisin more slowly, with 46.2% cumulative release at 48 h. Release curves differed between groups at multiple time points (p < 0.001), indicating that liposomal encapsulation delayed irisin release (Fig. 2C-D).

Fig. 2.

Evaluation of Encapsulation Efficiency and Sustained Release Performance of Lipo@Irisin. Note: (A) Schematic workflow of the encapsulation efficiency and in vitro sustained-release performance of T7@Lipo@Irisin; (B) Encapsulation efficiency of irisin in three independent batches of T7@Lipo@Irisin measured by LC-MS/MS; (C) Release profiles of Free-Irisin and T7@Lipo@Irisin in PBS supplemented with 10% FBS over 48 h; (D) Statistical comparison of release profiles between the two groups; (E) First-order kinetic model fitting of the release curves; (F) Histograms showing particle size distribution before and after release; (G) SEM images showing that liposomes retained a regular spherical morphology after 48 h of release

External medium was collected at various time points (2, 4, 8, 12, 24, and 48 h) to determine irisin concentrations and construct release kinetics curves. The release profile of T7@Lipo@Irisin was better fitted to a first-order kinetic model (R² = 0.973), likely due to the structural characteristics of the lipid bilayer (Fig. 2E). In addition, scanning electron microscopy (SEM) and particle size analysis before and after release revealed a slight increase in diameter from 168.2 nm to 179.5 nm, suggesting mild structural swelling due to hydration while overall maintaining good stability (Fig. 2F-G). Collectively, these data indicated that T7@Lipo@Irisin combined high encapsulation efficiency with sustained irisin release under serum-containing conditions.

Neuron-mimetic membrane coating enhances structural stability and surface exposure of T7 peptide

Raman spectroscopy was used to compare lipid ordering between T7@Lipo@Irisin and the biomimetic membrane-coated formulation (T7@Lipo@Irisin@Neuron Membrane). After membrane fusion, C-H symmetric stretching signals near 2850 cm⁻¹ and 2880 cm⁻¹ increased (Fig. 3 A), indicating increased lipid ordering within the bilayer and suggesting improved membrane stability. To determine whether the T7 peptide remained exposed on the nanoparticle surface after membrane fusion, a fluorescence quenching assay using FITC-T7 was conducted. In the pre-fusion sample, the addition of a quenching agent (Trypan Blue) resulted in a marked decrease in FITC fluorescence intensity, with a quenching efficiency of 62.3%. In contrast, post-fusion particles showed a significantly lower quenching efficiency of 11.7% (Fig. 3B). The change in quenching supported retention of surface-associated T7 after membrane coating and suggested a more ordered membrane environment that altered dye accessibility. Western blot analysis further confirmed that the fused nanoparticles retained key membrane proteins, including Na⁺/K⁺-ATPase and NCAM, verifying the structural integrity of the neuronal membrane coating (Fig. 3 C). TEM-based morphological analysis revealed a distinct membrane-like outer layer on the fused liposomes, accompanied by a slight increase in particle size while maintaining overall spherical uniformity (Fig. 3D).

Fig. 3.

Neuron Membrane Fusion Enhances Liposomal Order and Surface Exposure of T7 Peptides. Note: (A) Raman spectroscopy showing enhanced symmetric stretching peaks of lipid chains in neuron-mimetic liposomes, indicating improved membrane order; (B) FITC-T7 peptide quenching assay demonstrating that, after membrane fusion, T7 peptides remain exposed on the particle surface with high accessibility; (C) Western blot analysis confirming the presence of neuronal membrane markers Na⁺/K⁺-ATPase and NCAM in fused liposomes; (D) TEM image showing that neuron membrane-fused liposomes exhibit a well-defined outer membrane and regular morphology. **p < 0.01

Collectively, neuron-mimetic membrane coating improved bilayer ordering and structural stability and maintained measurable T7 surface presentation, supporting downstream targeting and cellular recognition.

Liposomes efficiently traverse the in vitro BBB model and are specifically taken up by neurons

To evaluate the ability of T7@Lipo@Irisin to cross the BBB and its specificity for neuronal uptake, an in vitro BBB co-culture model was established using a Transwell system, with hCMEC/D3 in the upper chamber and mouse neuroblastoma cells (Neuro-2a) in the lower chamber (Fig. 4 A). DiD-labeled liposomes were added to the apical chamber and incubated for 24 h. Confocal laser scanning microscopy (CLSM) was used to visualize liposome distribution in TfR+ endothelial cells and NeuN+ neurons. In the endothelial monolayer, T7@Lipo@Irisin produced stronger red fluorescence and higher colocalization with TfR staining than Lipo@Irisin, with an approximately 1.9-fold increase in fluorescence intensity (p < 0.001). In neurons in the lower chamber, the T7-modified group also showed enhanced red fluorescence that colocalized with NeuN-labeled cytoplasm, with a 2.6-fold increase in signal intensity (p < 0.001), indicating that the system efficiently crossed the BBB model and specifically targeted neurons (Fig. 4B-C).

Fig. 4.

Transcytosis and Neuronal Targeting of T7-Modified Liposomes in an In Vitro BBB Model. Note: (A) Schematic of the Transwell model: hCMEC/D3 endothelial cells were seeded in the upper chamber, and Neuro-2a neurons in the lower chamber; (B) Confocal microscopy showing the localization of Lipo@Irisin and T7@Lipo@Irisin in TfR⁺ endothelial cells and NeuN⁺ neurons; bar = 25 μm; (C) Quantification of fluorescence intensity: red indicates liposomes, green indicates NeuN, and blue indicates DAPI; (D) Flow cytometry analysis of liposome uptake in endothelial cells and neurons for both Lipo@Irisin and T7@Lipo@Irisin groups; (E) Bar graph summarizing flow cytometry data. *p < 0.05; ***p < 0.001. All cell-based experiments were repeated three times

Flow cytometry further quantified cellular uptake in both chambers. T7@Lipo@Irisin achieved uptake rates of 74.5 ± 3.2% in endothelial cells and 68.9 ± 2.8% in neurons, whereas Lipo@Irisin showed uptake rates of 39.4 ± 2.6% and 26.1 ± 3.5%, respectively (Fig. 4D-E). These findings demonstrate that T7@Lipo@Irisin possesses high transmembrane transport efficiency and strong neuronal targeting capability.

T7-modified liposomes significantly improve barrier integrity markers

Serum inflammatory cytokines and BDNF were measured in healthy donors and PND patients. IL-6 and TNF-α concentrations increased in the PND cohort, whereas BDNF decreased, indicating that PND patients exhibit a heightened inflammatory state (Fig. 5 A). Subsequently, to evaluate the protective effect of T7@Lipo@Irisin on BBB function in the PND model, a Transwell co-culture system was established and corresponding assays were performed (Fig. 5B). Following treatment with T7-modified liposomes, the TEER of the Transwell model significantly increased to 280.4 ± 8.7 Ω·cm², compared to 178.6 ± 7.3 Ω·cm² in the PND group (p < 0.001), indicating enhanced barrier tightness (Fig. 5 C). Consistent with the TEER findings, FITC-dextran flux decreased by 52%, indicating reduced paracellular permeability (Fig. 5D). Western blotting further showed increased expression of tight-junction proteins in brain microvascular endothelial cells after T7@Lipo@Irisin treatment. ZO-1 increased by 1.7-fold and Occludin increased by 2.3-fold relative to the PND group (both p < 0.01), supporting restoration of junctional components at the protein level (Fig. 5E-F). Together, these results underscore the potential of T7@Lipo@Irisin to preserve BBB integrity under pathological conditions.

Fig. 5.

In Vitro Validation of T7-Modified Liposomes in Enhancing the Structural and Functional Integrity of the BBB. Note: (A) ELISA analysis of serum IL-6, TNF-α, and BDNF levels in healthy donors (n = 10) and PND patients (n = 6); (B) Schematic diagram of the Transwell system used to construct the in vitro PND model; (C) TEER values measured in the Transwell model; (D) FITC-Dextran permeability assay; (E) Western blot bands showing ZO-1 and Occludin protein expression across groups; (F) Quantitative analysis corresponding to the Western blot results. ns, p > 0.05; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001

Improvement of mitochondrial membrane potential and ROS levels

Mitochondrial membrane potential was assessed by JC-1 staining. The PND group showed a reduced red-to-green fluorescence ratio, indicating mitochondrial depolarization. T7@Lipo@Irisin increased the ratio to 1.82 ± 0.13 compared with 0.94 ± 0.09 in the PND group, supporting recovery of mitochondrial membrane potential after treatment. Mitochondrial ROS was measured using MitoSOX. ROS-associated fluorescence increased in the PND group, whereas T7@Lipo@Irisin reduced signal intensity by approximately 40% relative to the PND group, indicating effective alleviation of oxidative stress and partial restoration of mitochondrial homeostasis (Fig. 6A-B).

Fig. 6.

T7@Lipo@Irisin Alleviates Neuronal Mitochondrial Dysfunction. Note: (A) JC-1 dual-staining fluorescence images and MitoSOX fluorescence images for the Sham, PND, Irisin, and T7@Lipo@Irisin groups; (B) Bar graphs showing JC-1 red/green fluorescence ratio and MitoSOX fluorescence intensity; (C) Bar graph of ATP luminescence intensity; (D-E) Western blot bands and corresponding quantitative analysis. ns, p > 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001

Cellular bioenergetic status was further evaluated by luciferase-based ATP measurement. ATP abundance in the T7@Lipo@Irisin group recovered to 95% of the Sham level and exceeded values measured in the PND and Free-Irisin groups, consistent with improved mitochondrial energy production (Fig. 6 C). Western blotting showed increased Bcl-2 and decreased Bax after T7@Lipo@Irisin treatment, resulting in a higher Bcl-2/Bax ratio than in other groups (p < 0.001). These results indicated that T7@Lipo@Irisin restored mitochondrial function and activated anti-apoptotic signaling pathways (Fig. 6D-E).

SiRNA interference confirms AMPK/PGC-1α dependency

To further elucidate the molecular mechanism underlying the neuroprotective effects of T7-modified liposomes, we constructed an AMPKα-specific siRNA interference model to block the upstream signaling pathway. First, PND model cells treated with Control siRNA were used to exclude potential effects of the transfection reagent, whereas AMPKα siRNA was applied to evaluate the protective role of AMPKα in PND. Mitochondrial membrane potential, ATP content, and gene-expression readouts did not differ between the PND group and the PND + control siRNA group (Fig. 7A–H), indicating that the transfection reagent did not affect experimental outcomes. In contrast, all indicators were markedly reduced in the PND + AMPKα siRNA group (Fig. 7A–H), suggesting that AMPKα expression exerts a protective effect in PND. Upon administration of AMPKα siRNA in the PND + T7@Lipo@Irisin group, mitochondrial membrane potential significantly decreased, approaching levels observed in the PND group (Fig. 7A-B). Similarly, ATP content dropped markedly and became comparable to that of the PND group (Fig. 7 C). Western blot analysis showed a substantial downregulation of p-AMPK, PGC-1α, and BDNF compared to the non-silenced group. RT-qPCR results confirmed significant reductions in PGC-1α and BDNF mRNA expression. These results indicate that the beneficial effects of T7-modified liposomes on mitochondrial function are dependent on activation of the AMPK/PGC-1α signaling pathway (Fig. 7D-H).

Fig. 7.

siRNA Knockdown Confirms the Neuroprotective Role of T7@Lipo@Irisin via the AMPK/PGC-1α Pathway. Note: Effects of T7@Lipo@Irisin on Neuro-2a PND model cells. (A) JC-1 fluorescence images (red indicates high membrane potential, green indicates low membrane potential); (B) Quantification of the red/green JC-1 fluorescence intensity ratio; (C) Bar graph showing ATP levels; (D) Western blot bands of p-AMPK, PGC-1α, and BDNF proteins, with GAPDH as a loading control; (E) Quantitative analysis of p-AMPK band intensity; (F) Quantitative analysis of PGC-1α band intensity; (G) Densitometry of BDNF; (H) RT-qPCR analysis of PGC-1α and BDNF mRNA levels. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001

Verification of pathway activation as a targeting mechanism

To further investigate whether the mitochondrial metabolic remodeling and antioxidative effects of T7@Lipo@Irisin were mediated through specific signaling pathways, Western blot and nuclear/cytoplasmic protein fractionation analyses were performed. Potential confounding effects from anesthesia and empty liposomes (T7@Lipo) on AMPKα expression and phosphorylation were not observed (Figure S1A–B). Control siRNA and AMPKα siRNA groups were then included to evaluate pathway dependence. Protein abundance did not differ between the PND and PND + control siRNA groups (Fig. 8A–E), indicating negligible influence from the transfection procedure. AMPKα siRNA reduced pathway-related protein levels (Fig. 8A–E), supporting a protective contribution from AMPKα in the PND context. Compared with the PND group, T7@Lipo@Irisin increased multiple energy-regulating and antioxidant-associated proteins. p-AMPK increased by approximately 2.3-fold, while PGC-1α, UCP2, SIRT1, and BDNF increased by roughly 2.9-, 3.6-, 4.1-, and 1.6-fold, respectively. Nrf2 increased by 4.6-fold (Fig. 8A–B). Consistent with these findings, RT-qPCR analysis revealed significantly higher mRNA levels of PGC-1α, UCP2, SIRT1, and BDNF in the T7@Lipo@Irisin group compared with the PND group (Fig. 8A–C). In addition, nuclear/cytoplasmic protein separation revealed a marked enhancement of Nrf2 nuclear translocation, with the nuclear-to-total Nrf2 ratio elevated nearly 3.2-fold. These findings suggest that the T7@Lipo@Irisin systemically activates the AMPK/PGC-1α/SIRT1/Nrf2 signaling axis, thereby enhancing neuronal anti-apoptotic and antioxidative capacity (Fig. 8D-E).

Fig. 8.

Verification of Mitochondrial Targeting Mechanism by Pathway Protein Expression Analysis of T7@Lipo@Irisin. Note: Gene and protein expression in hippocampal tissues of mice in each group. (A) Western blot bands showing the expression of p-AMPK, t-AMPK, PGC-1α, UCP2, SIRT1, Nrf2, and BDNF in the PND and T7@Lipo@Irisin groups; GAPDH was used as the internal control; (B) Quantitative bar graph of the grayscale intensity for the five proteins, demonstrating significantly higher expression levels in the T7@Lipo@Irisin group compared to the PND group; (C) RT-qPCR analysis of PGC-1α, UCP2, SIRT1, and BDNF mRNA levels; (D) Western blot analysis of nuclear and cytoplasmic fractions: Nrf2 expression was elevated in the nuclear fraction, indicating enhanced nuclear translocation; (E) Quantification of the Nrf2 nuclear translocation ratio by grayscale intensity; the T7@Lipo@Irisin group exhibited a significantly higher ratio than the PND group (p < 0.001, N = 6). **p< 0.01; ***p < 0.001; ****p < 0.0001

Enhanced metabolic flux and increased TCA cycle activity

To clarify the impact of T7@Lipo@Irisin on neuronal metabolic activity, a multi-parametric functional assessment was conducted. Seahorse metabolic analysis revealed that T7@Lipo@Irisin treatment significantly enhanced both oxidative phosphorylation and glycolytic function. Relative to the PND group, neurons in the T7@Lipo@Irisin group showed a 46% increase in basal OCR and a 38% increase in ECAR, indicating a marked upregulation in cellular energy metabolism (Fig. 9A-B).

Fig. 9.

T7@Lipo@Irisin Enhances Neuronal Metabolic Activity and TCA Cycle Intermediate Levels. Note: Metabolic measurements in hippocampal tissues across all experimental groups.(A) Seahorse analysis of basal OCR in neurons; (B) The ECAR, an indicator of glycolytic activity, was also significantly increased, suggesting that both mitochondrial function and glycolysis were activated; (C) Targeted metabolomic analysis showed that the relative levels of citrate, succinate, and α-ketoglutarate were markedly elevated in the T7@Lipo@Irisin group, indicating enhanced TCA cycle activity. **p < 0.01; ***p < 0.001

Targeted LC-MS/MS metabolomics further supported enhanced TCA cycle activity. Citrate, succinate, and α-ketoglutarate increased by 1.8-, 2.1-, and 1.9-fold, respectively, after T7@Lipo@Irisin treatment (Fig. 9 C). These results indicate that the liposome-mediated mitochondrial modulation effectively boosts neuronal metabolic flux and activates mitochondrial energy production pathways.

scRNA-seq reveals functional alterations in neuronal subpopulations in the disease model

In hippocampal tissues from the mouse disease model (Surgery) and the control group, scRNA-seq identified the major brain cell populations, including neurons, astrocytes, microglia, and endothelial cells (Fig. 10 A). Clustering visualization showed a marked reduction in neuronal subpopulations in the disease model, accompanied by significant transcriptional alterations.

Fig. 10.

scRNA-seq Reveals Altered Functional States of Neuronal Subpopulations in the Disease Model. Note: (A) UMAP clustering plot showing all identified cell subpopulations (neurons, astrocytes, microglia, endothelial cells, etc.); (B) Snap25 FeaturePlot illustrating the spatial distribution of the neuronal marker gene; (C-D) Comparison of the proportion of positive cells (% cells expressing > 0) between the control group (Control) and the Surgery group (Surgery) for Snap25 (C), Syt1, and Mki67 (D); (E) GO bubble plots of the control (Control) and Surgery groups; (F) Heatmap showing the transcriptional clustering of neuronal subpopulations across different groups

Snap25, a marker associated with mature neuronal identity, showed widespread and higher expression in the Control group, whereas Snap25 expression was markedly lower in the Surgery group (Fig. 10B). Quantitative analysis confirmed that the proportion of Snap25⁺ neurons declined from a high level in the control group to a much lower level in the disease model (Fig. 10 C), suggesting that central inflammation substantially suppressed the mature neuronal phenotype.

Differential expression and functional annotation indicated that Surgery-associated neuroinflammation increased expression of the mature synaptic marker Syt1 and reduced expression of the proliferation and neurogenesis marker Mki67 in hippocampal tissue. These findings indicated that the inflammatory environment might drive compensatory enhancement of mature neuronal synaptic function, while simultaneously inhibiting the generation of new neurons, thereby altering the functional state of neuronal populations (Fig. 10D). GO enrichment analysis demonstrated that the Snap25⁺ subpopulation in the control group was enriched in pathways such as “mitochondrial oxidative phosphorylation,” “ATP synthesis and nucleotide triphosphate metabolism,” and the “electron transport chain”. In contrast, the Surgery group was enriched for postoperative inflammation and immune activation signatures, including microglial activation pathways (Fig. 10E). Transcriptomic heatmaps further showed that the surgery group exhibited upregulation of microglia/immune-related cells and signaling molecules (e.g., Cd274/PD-L1), consistent with the GO immune enrichment results. Neuronal function-related modules were broadly attenuated, consistent with reduced energy-metabolism signatures (Fig. 10 F).

In summary, the single-cell transcriptomic analysis revealed a marked reduction in hippocampal neuronal subpopulations in the Surgery group, accompanied by substantial transcriptional reprogramming. The fraction of Snap25-positive neurons decreased relative to controls, consistent with impaired maintenance of mature neuronal programs and disrupted homeostasis. Increased Syt1 expression coupled with reduced Mki67 expression indicated concurrent strengthening of synaptic-associated transcriptional features and suppression of proliferation or neurogenesis-related signals within the same tissue context. At the pathway level, control samples were enriched for mitochondrial oxidative phosphorylation, electron transport chain activity, and ATP synthesis, whereas Surgery samples showed enrichment for immune and inflammatory activation, microglial activation, and elevated neuro-immune signaling, including Cd274 or PD-L1. Correspondingly, neuronal function and energy metabolism were globally weakened. Collectively, these results indicate that central inflammation suppresses neuronal maturation and neurogenesis while triggering synaptic remodeling through enhanced immune responses and disturbed mitochondrial metabolism, which may underlie inflammation-associated cognitive impairment and represent potential therapeutic targets.

ATAC-seq reveals enhanced chromatin accessibility and activation of metabolic regulation

scATAC-seq profiling of mouse brain tissue generated a chromatin accessibility map and resolved major cell populations, including neurons, astrocytes, microglia, endothelial cells, oligodendrocytes, and T cells (Fig. 11 A). UMAP clustering indicated reduced accessibility across neuronal populations in the Surgery group relative to controls, consistent with attenuated regulatory activity in disease-associated neuronal states.

Fig. 11.

scATAC-seq Reveals Chromatin Accessibility Decline and Impaired Metabolic Regulation in the Disease Model. Note: (A) UMAP plot showing chromatin accessibility-based clustering of cellular subpopulations; (B) IGV track of the Prkaa1 promoter region demonstrating increased chromatin accessibility; (C) IGV track of the Ppargc1a promoter region showing elevated chromatin peak intensity; (D) TSS enrichment plot (±2 kb) comparing genome-wide chromatin accessibility between groups; (E) Bar chart of motif enrichment illustrating the prevalence of metabolism-related transcription factors such as CREB1 and NRF1

Promoter accessibility differences were observed at key metabolic regulators. In the Control group, clear ATAC peaks were detected at the Prkaa1 (AMPKα) and Ppargc1a (PGC-1α) promoter regions (±2 kb). In the Surgery group, these peaks were markedly weakened or absent (Fig. 11B-C). Genome-wide transcription start site (TSS) enrichment analysis supported the promoter-level results, with a lower TSS enrichment profile in the Surgery group than in controls (Fig. 11D), indicating impaired promoter accessibility.

Motif enrichment analysis of high-accessibility regions revealed significant enrichment of binding sites for metabolism-related transcription factors, including Creb1, Nrf1, Sp1, and Esrra (Fig. 11E). Notably, the CREB1 motif was significantly enriched in the surgery group (p.adj = 0.0042), and NRF1 binding sites also showed statistical differences (p.adj = 0.0313). The combined motif shifts supported remodeling of transcription factor programs related to energy metabolism under surgery-associated stress and neuroinflammation.

In summary, scATAC-seq data indicated reduced chromatin accessibility at metabolic regulatory loci, decreased TSS enrichment, and altered motif landscapes in the Surgery group. The epigenetic pattern aligned with constrained downstream transcription and neuronal functional impairment and highlights regulatory nodes that may be leveraged to restore metabolic homeostasis.

Cicero and proteomics reanalysis reveal activation of long-range regulatory networks

Enhancer-promoter co-accessibility networks were inferred from scATAC-seq data using Cicero to evaluate contributions of distal regulatory elements during functional recovery. The number of inferred enhancer-promoter links increased under recovery-associated conditions (18,784 pairs) compared with the disease-associated state (7,242 pairs), indicating broader chromatin reconfiguration accompanied by expansion of putative regulatory connectivity (Fig. 12A-C).

Fig. 12.

Enhancer-Promoter Interactions and Proteomic Data Reveal Activation of Neurotrophic and Metabolic Regulatory Networks. Note: (A-B) Cicero co-accessibility analysis showing enhancer-promoter interactions at the Sirt1 (A) and Nfe2l2 (B) loci in the disease (Disease) and recovery (Recovery) groups, with purple arcs representing high-confidence chromatin connections (co-accessibility score > 0.35); (C) Comparison of the total number of enhancer-promoter interactions between the disease and recovery groups; (D) GO functional enrichment of genes associated with newly formed interactions in the recovery group, highlighting significant involvement in mitochondrial metabolism and synapse-related pathways; (E) Volcano plot of differentially expressed proteins between the disease and recovery groups, with key upregulated proteins (SIRT1, NDUFS2, BDNF) labeled; (F) STRING PPI network analysis showing that the upregulated proteins formed a highly connected regulatory module encompassing mitochondrial biogenesis, synaptic structural maintenance, and oxidative stress responses

At the Sirt1 locus, consistent co-accessibility was detected between distal enhancers and the promoter, and the linked enhancer regions overlapped with features consistent with active regulatory potential (Fig. 12B). Similarly, around the oxidative stress regulator Nfe2l2 (Nrf2), multiple distal enhancer-promoter connections were detected in the recovery group, indicating stronger epigenetic activation under recovery conditions (Fig. 12 C).

Genes linked to newly detected or strengthened interactions were enriched for pathways related to mitochondrial maintenance, NAD-dependent metabolism, ROS clearance, and synaptic homeostasis (Fig. 12D), indicating an association between network remodeling and recovery-relevant functional programs.

Consistent with the epigenetic results, TMT-based proteomics identified 286 upregulated proteins in the recovery group (|log₂FC| > 0.58, p < 0.05), including key metabolic and neurotrophic factors such as SIRT1, NDUFS2, and BDNF (Fig. 12E). STRING protein-protein interaction analysis showed that the upregulated proteins formed a connected module enriched for mitochondrial biogenesis, synaptic maintenance, and oxidative stress responses (Fig. 12 F).

Collectively, integrated scATAC-seq and proteomic analyses supported increased enhancer-promoter connectivity during recovery, with coordinated activation of mitochondrial and metabolic programs at both chromatin and protein levels.

T7@Lipo@Irisin improves mitochondrial dysfunction in the brain tissue of PND mice

To evaluate in vivo brain distribution and biosafety, DiR-labeled Lipo@Irisin and T7@Lipo@Irisin were administered by tail vein injection, followed by near-infrared IVIS imaging and toxicity assessment. Both formulations produced detectable fluorescence in major organs, whereas T7@Lipo@Irisin produced stronger signals in the brain region (Figure S2A). Brain fluorescence increased over time at 1 h, 6 h, and 24 h, supporting time-dependent enrichment in brain tissue (Figure S2B). Quantitative Region-of-interest (ROI) analysis showed that at 24 h post-injection, the fluorescence signal intensity in the brain of the T7-modified group was 2.3-fold higher than that of the unmodified group (Figure S2C). Pretreatment with TFR1 antibody significantly suppressed this accumulation (Figure S2A–C), indicating that the enhanced brain enrichment of T7-modified liposomes results from specific central targeting rather than nonspecific retention. To further verify the biosafety of the delivery system, systemic toxicity evaluation was performed on postoperative days 7 and 14. MicroCT imaging revealed no significant differences in the structure or volume of major organs such as the liver, kidney, and lung between the T7-modified and control groups (Figure S2D). H&E staining showed no necrosis, hemorrhage, or inflammatory infiltration in the heart, liver, lung, or kidney, and overall tissue structures remained intact, indicating good tissue compatibility of the liposomal system (Figure S2E). Serum biochemistry further showed ALT, AST, BUN, Cr, and C3a values within normal ranges, with no significant intergroup differences (Figure S2 F). Collectively, these findings demonstrate that T7@Lipo@Irisin enables stable and targeted delivery to brain tissue without inducing significant systemic toxicity.

Mitochondrial function in brain tissue was evaluated on postoperative day 7. JC-1 staining showed a reduced red to green fluorescence ratio in the PND group relative to the Sham group, indicating mitochondrial depolarization. Lipo@Irisin produced a modest increase, whereas T7@Lipo@Irisin increased the ratio and approached the Sham level, with significant intergroup differences. Fluorescence imaging supported the quantitative results and showed stronger JC-1 aggregate signals in the T7@Lipo@Irisin group (Fig. 13A-B). MitoSOX staining showed elevated mitochondrial ROS in the PND group, reaching approximately 2.5-fold of the Sham level. Lipo@Irisin reduced ROS partially, whereas T7@Lipo@Irisin reduced ROS to approximately 1.2-fold of the Sham level. Fluorescence images showed intense red signal accumulation in the PND group that was reduced after T7@Lipo@Irisin treatment, consistent with attenuation of oxidative stress (Fig. 13C-D). To further evaluate mitochondrial permeability, calcium-induced mPTP opening was assessed. The PND group displayed a rapid decline in OD₅₄₀ absorbance, indicating a significantly increased mPTP opening rate compared to the Sham group. The PND + Lipo@Irisin group showed a modest alleviation, whereas the OD₅₄₀ decline curve in the PND + T7@Lipo@Irisin group was substantially flatter, with the opening rate nearly restored to baseline levels, suggesting that the intervention effectively inhibited abnormal mPTP opening and improved mitochondrial stability (Fig. 13E). Western blotting was used to assess the tight junction proteins ZO-1 and Occludin, which contribute to BBB structure. Protein abundance decreased in the PND group and increased after T7@Lipo@Irisin treatment, with values approaching the Sham level, consistent with preservation of BBB structural integrity (Fig. 13F-G). In PND mice treated with Free-Irisin, mitochondrial function and BBB structural integrity showed partial improvement compared with the Sham group. However, the therapeutic effects were substantially weaker than those observed in the Lipo@Irisin and T7@Lipo@Irisin groups, suggesting that Free-Irisin alone has limited efficacy in restoring mitochondrial function and BBB integrity (Fig. 13A-G). Furthermore, AMPK siRNA treatment markedly suppressed the recovery of mitochondrial function and BBB integrity induced by T7@Lipo@Irisin in PND mice (Fig. 13A–G), indicating that the therapeutic effect of T7@Lipo@Irisin is likely mediated through AMPK-related signaling pathways.

Fig. 13.

T7@Lipo@Irisin Restores Mitochondrial Function in the Brains of PND Mice. Note:(A) JC-1 dual staining images; bar = 25 μm; (B) Quantification of red-to-green fluorescence ratio; (C) MitoSOX staining images of mitochondrial ROS; (D) Quantitative analysis of fluorescence intensity; bar = 25 μm; (E) Absorbance curve of OD540 indicating mPTP opening dynamics; (F-G) Western blot analysis showing expression levels of ZO-1 and Occludin in brain tissue, with GAPDH as the loading control and corresponding quantification. N = 6. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001

MWM test confirms the effect of T7@Lipo@Irisin on improving cognitive function, working memory, and recognition behavior

To evaluate the impact of T7@Lipo@Irisin on learning and memory in PND mice, the MWM test was conducted on postoperative day 7 (Fig. 14 A). Mice in the PND group showed longer escape latency across training, consistent with impaired spatial learning. T7@Lipo@Irisin reduced escape latency over training sessions, indicating improved acquisition of the task (Fig. 14B). In the subsequent spatial probe test, the T7@Lipo@Irisin group crossed the former platform location 4.8 ± 0.6 times, indicating substantial restoration of spatial memory (Fig. 14 C). Representative swimming trajectory plots further illustrated that mice in the T7@Lipo@Irisin group concentrated their search within the target quadrant (Fig. 14D). To further assess the recovery of short-term memory and recognition function, Y-maze spontaneous alternation and NOR tests were performed. In the Y-maze test, mice treated with T7@Lipo@Irisin showed a significantly increased spontaneous alternation rate of 78.3% (vs. 52.9% in the PND group, p < 0.001), approaching the Sham level (Fig. 14E). In the NOR test, the T7@Lipo@Irisin group spent significantly more time exploring the novel object, with a recognition index of 0.69 ± 0.05 (vs. 0.41 ± 0.04 in the PND group, p < 0.001), reflecting effective recovery of object recognition ability (Fig. 14F-H). These behavioral findings collectively demonstrate that systemic intervention with liposomes exerts a significant therapeutic effect in enhancing perioperative neuronal function. Free-Irisin produced modest improvements in cognitive performance, working memory, and recognition behavior relative to the PND group, but the magnitude of benefit remained lower than that observed with Lipo@Irisin and T7@Lipo@Irisin, indicating that liposomal encapsulation enhanced irisin efficacy under the conditions tested (Fig. 14A-H). AMPK siRNA administration significantly attenuated the improvements in cognitive function, working memory, and recognition performance induced by T7@Lipo@Irisin (Fig. 14A–H), further indicating that T7@Lipo@Irisin exerts its therapeutic effects through AMPK-dependent pathways.

Fig. 14.

Morris Water Maze Behavioral Test Demonstrates Cognitive Improvement in PND Mice Treated with T7@Lipo@Irisin. Note:(A) Experimental workflow: the Morris water maze test consisted of a habituation period, a training phase, and a spatial probe trial; (B) Learning curve showing escape latency during the training period; (C) Number of platform crossings during the spatial probe trial; (D) Representative swimming paths of mice from each group; (E) Bar graph of Y-maze spontaneous alternation rates; (F) Schematic of the NOR test; (G) Illustration of the placement of familiar and novel objects as well as representative exploration trajectories of mice in each group; (H) Bar graph of the NOR index. N = 12, *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001

Neurological scores and histopathological improvements were significantly enhanced

Neurological function and brain pathology were evaluated on postoperative day 7. Garcia scoring showed higher neurological performance in the T7@Lipo@Irisin group (14.8 ± 0.6) than in the PND group (10.2 ± 0.7, p < 0.001), with scores approaching the Sham level (Fig. 15 A). Nissl staining of the hippocampus showed that neurons in the T7@Lipo@Irisin group were densely and orderly arranged, with well-defined cytoplasmic staining and intact cellular structures. In contrast, the PND group exhibited disorganized cell arrangement and cytoplasmic vacuolization, both of which were notably improved by the intervention (Fig. 15B). To further assess neuronal apoptosis, immunofluorescence staining for cleaved caspase-3 was performed. The number of cleaved caspase-3-positive cells in the T7@Lipo@Irisin group was reduced by approximately 50%, indicating a significant suppression of postoperative neuronal apoptosis (Fig. 15 C). These findings suggest that T7@Lipo@Irisin facilitates functional recovery following perioperative neuronal injury by alleviating apoptosis and improving pathological morphology. Western blot analysis further confirmed the anti-apoptotic effects of T7@Lipo@Irisin (Fig. 15D). In the PND group, the pro-apoptotic protein Bax was markedly upregulated, whereas the level of the anti-apoptotic protein Bcl-2 was decreased. T7@Lipo@Irisin treatment effectively restored the Bcl-2/Bax ratio, thereby inhibiting neuronal apoptosis.

Fig. 15.

Behavioral and Histopathological Analyses Reveal Significant Neuroprotection by T7@Lipo@Irisin in PND. Note:(A) Garcia neurological score results (N = 12); (B) Nissl staining of brain sections; bar = 50 μm; (C) Immunofluorescence staining of Cleaved Caspase-3 in hippocampal tissues (scale bar = 25 μm); (D) Western blot analysis and quantification of Bcl-2/Bax expression in hippocampal tissues (N = 6; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001).

Similarly, compared with the PND group, the Garcia neurological scores of mice in the PND + Free-Irisin group increased but did not reach statistical significance (Fig. 15 A). Histopathology revealed partial improvement, with reduced neuronal apoptosis, but the therapeutic effects remained markedly weaker than those observed in the Lipo@Irisin and T7@Lipo@Irisin groups, again demonstrating the superior efficacy conferred by liposomal delivery (Fig. 15A-D). Compared with the T7@Lipo@Irisin group, all measured indicators were significantly decreased in the T7@Lipo@Irisin + AMPK siRNA group (Fig. 15A–D), indicating suppression of the therapeutic effects. These results suggest that the neuroprotective actions of T7@Lipo@Irisin are dependent on AMPK-related signaling pathways.

Discussion

In this study, a biomimetic liposomal platform modified with T7 peptide and fused with neuronal cell membranes was constructed to deliver irisin (T7@Lipo@Irisin). Evidence from in vitro BBB models and in vivo imaging supported BBB translocation and brain enrichment relative to Lipo@Irisin and Free-Irisin. Multi-omics analyses, including scRNA-seq, chromatin accessibility profiling, and proteomics, linked T7@Lipo@Irisin treatment to activation of the AMPK/PGC-1α axis, improved mitochondrial bioenergetics, altered neuron and glia interaction programs, reduced oxidative stress, and improved cognitive performance in the PND model. The integrated results provide mechanistic support for metabolism-targeted intervention strategies in PND.

From a drug delivery perspective, the T7@Lipo@Irisin system incorporates T7 peptide and neuronal membrane–derived components onto a liposomal scaffold, thereby enhancing irisin delivery efficiency and improving its therapeutic efficacy. Brain-targeting nanocarriers have commonly relied on ligands such as RVG, Angiopep-2, or anti-TfR antibodies [28]. T7 peptide provides a chemically defined alternative with controllable synthesis and lower immunogenic risk, and preclinical studies have supported its utility in models of stroke and hydrocephalus [29, 30]. Biomimetic delivery strategies based on cell membranes have also expanded options for modulating biodistribution and cellular interactions [31, 32]. Previous work has shown that liposomes functionalized with NK cell membrane proteins can selectively target MPP⁺-injured neurons while reducing macrophage uptake, enabling therapeutic delivery to treat Parkinson’s disease via meningeal lymphatic pathways [33]. In the present study, we employed neuronal membrane fusion to further enhance the neuro-affinity and membrane fusion efficiency of the liposomal system. Fluorescent tracing experiments and 3D Transwell assays demonstrated robust barrier penetration capacity and cellular targeting specificity, suggesting promising potential for the application of this delivery strategy in neurological disorders.

Our results also expand the understanding of irisin’s neuroprotective potential. Although recent studies have confirmed the neuroprotective role of irisin in models of Alzheimer’s disease, cerebral ischemia, and stress-related cognitive disorders [34–36]. However, much of the existing evidence has relied on peripheral administration or in vitro systems, and delivery optimization and pathway-level validation have often remained limited. Encapsulation of irisin in a biomimetic nanodelivery platform enabled evaluation of brain-directed delivery together with mechanism mapping, and results identified the AMPK/PGC-1α axis as a core mediator of irisin-associated neuroprotection in CNS injury.

At the level of mitochondrial function, T7@Lipo@Irisin significantly restored neuronal membrane potential, increased ATP production, and suppressed both ROS generation and mPTP opening in the PND model. These effects suggest that the intervention effectively counteracted mitochondrial dysfunction induced by perioperative stress. Conventional mitochondria-targeted antioxidant strategies, including MitoQ and SS31, primarily reduce oxidative injury through ROS scavenging [37, 38]. In contrast, T7@Lipo@Irisin improved mitochondrial function through AMPK-mediated metabolic reprogramming. Seahorse profiling and targeted metabolomics further supported recovery of respiratory capacity together with increased TCA cycle intermediates, indicating strengthened mitochondrial energy production through coordinated activation of metabolic pathways.

The AMPK/PGC-1α signaling axis is widely recognized as a regulator of energy sensing and mitochondrial biogenesis, whereas systematic evidence defining its mechanistic contribution to PND has remained limited [13, 15]. In this study, we utilized both siRNA- and CRISPR-based perturbations in cellular and animal models to validate the essential contribution of this pathway to the therapeutic effects of irisin-loaded liposomes. Measurement of p-AMPK, PGC-1α, SIRT1, and UCP2 further defined downstream molecular responses consistent with pathway engagement during mitigation of PND-associated injury.

Single-cell multi-omics analyses were applied to characterize PND-associated cellular states and to define mechanisms linked to T7@Lipo@Irisin intervention. For scRNA-seq, cell-type annotation was performed with reference to the hippocampal single-cell atlas reported by Suo et al., and results indicated recovery of neuronal and microglial subpopulation features after treatment, together with reduced inflammatory signaling in astrocytes. scATAC-seq and Cicero analyses identified increased enhancer-promoter connectivity at loci related to AMPK, PGC-1α, and mitochondrial metabolism genes, with enrichment in energy-metabolism and synaptic homeostasis pathways. Consistent with these findings, differential proteomics integrated with STRING network mapping showed coordinated changes in mitochondrial respiratory chain complexes and neurotrophic factors within a connected interaction module. Across transcriptomic, chromatin, and protein-interaction layers, the multi-omics results converged on AMPK/PGC-1α-centered regulation as a cross-level mechanistic framework for T7@Lipo@Irisin-associated neuroprotection (Fig. 16).

Fig. 16.

Integrated Multi-Omics Mechanistic Model. Note: This model integrates multi-omics evidence (single-cell RNA sequencing, single-cell ATAC sequencing, and proteomics) along with functional assays to illustrate how T7@Lipo@Irisin activates the AMPK/PGC-1α axis to restore mitochondrial homeostasis and neuro–glial interactions in postoperative neurocognitive disorder

Systemic biosafety was also evaluated. At the administered dose, T7@Lipo@Irisin did not produce detectable abnormalities in hepatic or renal function indices, major organ morphology, or blood biochemical parameters, supporting acceptable in vivo biocompatibility under the tested conditions. Neuronal membrane coating may reduce immune recognition and clearance and may contribute to prolonged circulation and improved targeting efficiency.

Despite the substantial progress made in this study, several limitations remain. First, experiments relied on a single cellular model and one PND animal model, and the therapeutic performance of T7@Lipo@Irisin has not been tested across models that vary by age, sex, or anesthetic regimen. Future work should evaluate additional models to define generalizability. Translation to humans also faces challenges related to BBB heterogeneity, inter-individual health status, and perioperative variability. Expanded clinical sampling through collaborative studies will be needed to strengthen translational relevance. Second, the current study focused on short-term intervention and observation. The stability and immunogenicity of the liposomal formulation under long-term or chronic pathological conditions still require systematic evaluation. As an essential component of preclinical development, future work will also involve comprehensive pharmacokinetic profiling of T7@Lipo@Irisin [39]. Furthermore, although we demonstrated that the AMPK/PGC-1α signaling axis plays a key role in mediating the protective effects of T7@Lipo@Irisin in PND, AMPK knockdown did not completely abolish its therapeutic efficacy. This suggests that additional signaling pathways may also contribute and provides a valuable direction for subsequent mechanistic exploration.

In conclusion, this study developed and validated a biomimetic liposomal system, T7@Lipo@Irisin, with robust brain-targeting capability, which markedly ameliorated mitochondrial dysfunction and neurobehavioral deficits in a perioperative neurocognitive disorder model. Integrated multi-omics analyses further highlighted the central role of the AMPK/PGC-1α axis in mediating the therapeutic actions of irisin in PND, offering a conceptual foundation and new insights for the application of irisin-based interventions.

Abbreviations

ALT

Alanine aminotransferase

AST

Aspartate aminotransferase

ATP

Adenosine triphosphate

AUC

Area under the curve

BCA

Bicinchoninic acid

BBB

Blood-brain barrier

BUN

Blood urea nitrogen

CLSM

Confocal laser scanning microscopy

Cr

Creatinine

DI

Discrimination index

DLS

Dynamic light scattering

DiR

1,1’-dioctadecyl-3,3,3’,3’-tetramethylindotricarbocyanine iodide

ECAR

Extracellular acidification rate

ECL

Enhanced chemiluminescence

ELISA

Enzyme-linked immunosorbent assay

FITC-T7

Fitc-labeled T7 peptide

FTIR

Fourier-transform infrared spectroscopy

GEO

Gene expression omnibus

GO

Gene ontology

H&E

Hematoxylin and eosin

IHC

Immunohistochemistry

KEGG

Kyoto encyclopedia of genes and genomes

LC-MS/MS

Liquid chromatography-tandem mass spectrometry

Lipo@Irisin

Irisin-loaded liposomes

LSI

Latent semantic indexing

MFI

Mean fluorescence intensity

mPTP

Mitochondrial permeability transition pore

NOR

Novel object recognition

OCR

Oxygen consumption rate

PDI

Polydispersity index

PND

Perioperative neurocognitive disorders

PPI

Protein-protein interaction

qPCR

Quantitative PCR

ROS

Reactive oxygen species

scATAC-seq

Single-cell assay for transposase-accessible chromatin using sequencing

scRNA-seq

Single-cell RNA sequencing

SEM

Scanning electron microscopy

T7@Lipo@Irisin

T7 peptide-modified liposomes encapsulating irisin

TEM

Transmission electron microscopy

TEER

Transepithelial electrical resistance

TfR

Transferrin receptor

TCA

Tricarboxylic acid

TMT

Tandem mass tag

TSS

Transcription start sites

Author contributions

Huisheng Wu and Wenlong Dai conceived and designed the study. Huisheng Wu, Wenlong Dai, Jun Cheng, and Aowen Li performed the experiments and collected the data. Yue Peng contributed to data analysis and interpretation, including multi-omic analyses. Jun Cheng and Aowen Li assisted with animal modeling and behavioral assessments. Wenlong Dai and Yue Peng carried out the bioinformatics and proteomics analyses. Huisheng Wu drafted the manuscript. Peipei Guo and Zhaohong Kong supervised the project, provided critical revisions, and acquired funding. All authors reviewed and approved the final manuscript.

Funding

This study was supported by the Research Fund of Anhui Institute of translational medicine (grant number: 2023zhyx-C61), the Research Fund Project of Anhui Medical University (grant number No 2022xkj140,2022xkj148), Science Research Fund of Colleges and University in Anhui Province, China (Grant No. 2022AH051187), Health Commission of Hubei Province scientific research project (No. WJ2021M143), The Fundamental Research Funds for the Central Universities (No. 413000714), The General scientific research projects of the Hubei Province Pathophysiology Society (No.2025HBAP013).

Data availability

All data generated or analyzed during this study are included in this article and/or its supplementary material files. Further enquiries can be directed to the corresponding author.

Declarations

Ethics approval and consent to participate

The study was approved by the Animal Ethics Committee of the First Affiliated Hospital of Anhui Medical University. Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.