Dysregulation of neurovascular unit in the retina after optic nerve injury Authors: Qiong Wu, Hui Wang, HongJuan Liu, Luyin Zhang, Qiping Wei Published: 2026 ABSTRACT PurposeTo investigate the changes in the neurovascular unit (NVU) of the retina in rats following optic nerve (ON) injury, and to explore the translational implications for traumatic optic neuropathy (TON).MethodsThe ON transverse quantitative traction (ONTQT) was performed to establish the model of ON and retinal injury. The rats were divided into the sham operation group (SG) and the model group (MG). At 14th day post-modeling, flash visual evoked potential (FVEP) test was performed to evaluate the visual function. Transmission electron microscopy (TEM) was used to observe the microstructure of retinal NVU. RNA binding protein with multiple splicing (RBPMS) immunofluorescence was applied to detect the survival retinal ganglion cell (RGC). The activity of astrocytes and Müller cells in retina was detected by glial fibrillary acidic protein (GFAP) immunofluorescence. The expression of tight junction proteins (Claudin-1, Claudin-5) and glial end feet markers aquaporin-4 (AQP4) and inwardly rectifying potassium channel subtype 4.1 (Kir4.1) in retinal tissue were test by western blot and Quantitative Real-Time Polymerase Chain Reaction (qRT-PCR).ResultsAt 14th day following ONTQT, the FVEP results exhibited the prolonged peak latency of P2 and the reduced amplitudes of N1-P1 and N2-P2. TEM showed structural changes of the basement membranes in NVU and ultrastructural abnormalities of tight junctions (TJs) after ONTQT. Besides, the expression of RBPMS in ganglion cell layer (GCL) was down-regulated and GFAP was over-expression in the injured retinal sections. The relative expressions of claudin-1and claudin-5 declined and the mRNA levels of AQP4 increased in the retina at 14 days following ONTQT. The mRNA levels of Kir4.1 was downregulated in the retina of MG.ConclusionsONTQT can be applied in the model of ON and retina injury. The dysfunction of retinal NVU may promotes the optic degeneration in rats following ONTQT, contributing to the RGC loss and impaired visual function. These findings provide a mechanistic basis for NVU-targeted neuroprotection and identify potential clinical biomarkers for the diagnosis and treatment of TON.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12886-026-04886-7. Introduction In retina, the blood-retinal barrier (BRB) system can be divided into inner blood-retina barrier (iBRB) and outer blood-retina barrier (oBRB). The iBRB comprises endothelial cells (ECs) that forming a single-layer lining of the retinal vasculature, which rests on a basal lamina that is covered by the processes of astrocytes and Müller cells [1]. The iBRB regulates the paracellular movement of fluids and molecules across retinal capillaries and plays a fundamental role in the microenvironment of the neural retina. While the oBRB is formed by the tight junctions (Tjs) between retinal pigment epithelial (RPE) cells, which rests on underlying Bruch’s membrane and regulates the paracellular movement of fluids and molecules between the choriocapillaris and the retina [2]. The general make-up of the iBRB consists of the neurovascular unit (NVU), in retina which is a complex multi-cellular structure composed of retina neurones, retinal vascular ECs with their dual basement membrane, surrounded by pericytes (PCs) and glial cells including astrocytes, Müller cells and microglia [3]. The retina relies on the NVU as a structural and functional unit to regulate blood flow and energy such as oxygen and glucose in the various pathophysiologic environment. The non-neuronal cells in the NVU are critical for the survival and homeostasis of neurons and maintenance of their functions. Increasing evidence demonstrated that the dysfunction of NVU among vascular components, neurons and glial cells can lead to many neurodegenerative and neuroinflammatory diseases, such as stroke, Alzheimer’s disease, Parkinson’s disease, and optic neuropathies [4]. As a crucial component of NVU, astrocytes can regulate the iBRB by releasing trophic factors, proinflammatory cytikines and shifting molecules structures of TJs [5]. The research suggested that in the early stage of diabetic retinopathy, the disruption of NVU occurs, leading to the decrease number of astrocytes and the impairment in the structure and function of microvasculature and neurons [6]. In glaucoma and traumatic optic neuropathy (TON), the breakdown of BRB plays a pivotal role in driving retinal neurodegeneration, optic nerve (ON) axon degeneration and retinal ganglion cell (RGC) death [4, 7]. TON is a rare disease that can lead to severe irreversible visual impairment. However, there is still no reliable evidence for treatment of TON with corticosteroids, surgical optic canal decompression or both confers any benefit [8]. Effective clinical targets for the treatment of TON remain elusive and the specific mechanisms of NVU dysfunction following ON injury remain incompletely understood. Therefore, it makes sense to investigate the mechanism of NVU in TON and boost the development of effective clinical interventions. TON occurs usually secondary to orbital, ocular, head, or traumatic facial injuries. TON occurs in up to 5% of all closed head injuries and 2.5% of maxillofacial and mid-face traumas [9]. Indirect TON exhibits a higher prevalence compared to direct TON. In general, TON is characterized by a loss of vision associated with RGC death and axonal loss of the ON [10]. Three classical rat models are widely applied in studying retinal and ON injury, which were ON transection, ON crush (ONC), and ocular blast injury [11]. Among them, ONC and ocular blast injury models are both models of indirect TON. Instead of ocular blast injury model, ONC model could avoid serious anterior and posterior segment damage and decrease mortality rate of animal. However, the drawback of the ONC model though is that it cannot calculate the force applied on the forceps to obtain crush injury [12]. Therefore, we improved the ONC model and developed a new model of ON transverse quantitative traction (ONTQT), which could both simulate model of indirect TON and calculate the force applied on the ON by transverse quantitative traction. In previous studies, we confirmed that the survival rates of RGCs at 1, 3, 7 and 14 days post-modeling were 78.94%, 60.07%, 38.92%, and 17.31% respectively. The modeling was achieved by ON transverse traction horizontally for a duration of 20 s (S) with a force of 0.1 N (N) [13]. Previous research has shown that the integrity of the NVU upon ONC was destroyed in the retina of mice and MMP-3 can attenuate neuroinflammation and neurodegeneration by tightening the glia limitans [4]. Therefore, in this study we aimed to investigate the complex interplay between RGC, astrocytes and microvascular cells in a new ON injury model of ONTQT and provide a mechanistic basis for the diagnosis and treatment of TON. The study’s hypothesis is that ONTQT causes the NVU dysfunction of inner retina, including the impairment of two basement membrane constitutes of astrocyte end-feet and ECs/PCs, which accelerate the RGC loss. Following experimental TON, we investigated changes in the retinal iBRB and NVU function through ultrastructural observation and evaluation of tight junction (TJ) proteins and astrocyte end-feet proteins. Elucidating the pathophysiological sequence following ON injury is crucial for optimizing clinical management and developing novel strategies to prevent or reverse vison loss in patients with TON. Methods Animals All animal procedures were approved by the Animal Ethics and Welfare Committee of Beijing Tongren Hospital Affiliated to Capital Medical University (Approval No.: TRLAWEC2022-S169). Specific pathogen-free (SPF) male Wistar rats aged 6 to 8 weeks were obtained from Beijing Vital River Laboratory Animal Technology Co., Ltd. A total of 24 rats were housed in a SPF animal facility under controlled conditions (22 °C constant temperature, 12 h light/12 h dark cycle) with free access to food and water. All animal experimental procedures were strictly in accordance with the national guidelines for the care and use of laboratory animals, including the Guide for the Care and Use of Laboratory Animals and the Beijing Laboratory Animal Welfare and Ethics Review Criteria issued by the Ministry of Science and Technology of China. ONTQT surgery Before model establishment, funduscopic examination was performed using an ophthalmoscope (Olympus Co., Ltd., Tokyo, Japan) to verify the integrity of the retina and its vasculature. The ONTQT model was established referring to a previously reported method with minor modifications [12]. Briefly, following anesthesia with 2% pentobarbital sodium (50 mg/kg, i.p.), each rat was positioned prone. The surgical field was sterilized with iodine, and a lateral canthal incision was made in the left orbit using conjunctival scissors. The oblique rectus muscle was carefully detached to expose the ON, which was then bluntly dissected from its sheath (Fig. 1A). For the model group (MG), a 6 − 0 polyester suture was looped around the ON at 2 mm posterior to the globe. The loop’s circumference was matched to that of the ON to avoid constriction. A digital force gauge (DNT-50, TECLOCK, Japan) was attached to the suture end, and a perpendicular traction force of 0.20 N was applied horizontally for 20 s (Fig. 1B). Post-operatively, the skin incision was closed, and the wound was disinfected with iodine and treated with topical erythromycin ointment. After model establishment, rats were carefully evaluated. Only those presenting with mydriasis, absent direct pupillary light reflex, preserved consensual light reflex, no evidence of retinal vascular occlusion, and transient retinal ischemia lasting less than 5 min were included in the study. In the sham-operated control group (SG), the ON was surgically exposed but not subjected to traction. Fig. 1ONTQT modeling. A. The exposed optic nerve (black arrow) B. A representative image of ON transverse quantitative traction surgery ONTQT modeling. A. The exposed optic nerve (black arrow) B. A representative image of ON transverse quantitative traction surgery Flash visual evoked potentials (FVEP) test The rats underwent FVEP testing 14 days post-injury. After being adapted to the dark for 15 min, the rats were anesthetized and fixed to the experiment station. The waveforms for each rat were recorded using an electrophysiological tester (Roland Electronics Co., Ltd., Keltern, Germany). A silver electrode was implanted beneath the skin at the midpoint of the line connecting the ear roots, which was recording electrode. The reference electrode was implanted into the midpoint of the binoculus and the ground electrode was inserted into the tail. When recording from the left eye, the right eye was covered with a self-made black eyeshade. White flash stimuli were delivered 50 times at frequency of 1 Hz (Hz) for a duration of 250 ms. To evaluate visual function, the latency (ms) and amplitude (µv) of the P1 and P2 waves in rats from each group were compared [14, 15]. The FVEP test of every rat were repeated four times and there were 4 rats in each group received FVEP test. After FVEP test, all rats were euthanized by an intraperitoneal overdose of pentobarbital sodium (150 mg/kg). Death was verified by respiratory and cardiac arrest before tissue collection. All procedures were performed in accordance with the institutional guidelines for animal welfare. Immunofluorescence protocol The eyeball tissue was harvested and fixed immediately in 10% formalin. The tissue was subsequently dehydrated in graded ethanol series and embedded in paraffin. Four-µm serial paraffin sections were cut along the vertical meridian with a microtome (CR-601ST, Jinhua Craftek lnstrument Co., Ltd., JinHua, China). Tissues were washed in PBS (3 × 5 min) and then incubated in blocking solution (10% goat serum) for 0.5 h at room temperature, followed by incubation in primary antibodies either overnight at 4 °C. The primary antibodies used were: glial fibrillary acidic protein (GFAP) monoclonal antibody (1:200, 60190-1-lg, Proteintech), RNA binding protein with multiple splicing (RBPMS) polyclonal antibody (1:200, 15187-1-AP, Proteintech). The next day, tissues were washed in PBS (3 × 5 min) and incubated with secondary antibodies conjugated to PE (1:200, goat anti–rabbit; Servicebio) for 1 h. Retinas were incubated with the nuclear dye DAPI for 20 min, washed in PBS (3 × 5 min), before microscopic analysis. All sections were observed under fluorescence microscope (Nippon Kogaku, ELCIPSE-CI) and x 200, x 400 magnification photomicrograph images were captured. The results were analyzed using an imaging analysis system. Western blot analysis Retinal tissues were carefully dissected on ice and immediately lysed in RIPA lysis buffer containing a protease inhibitor cocktail. The lysates were centrifuged at 20,000 rpm for 15 min at 4 °C (Scilogex, LLC, Rocky Hill, CT, USA), and the supernatant was collected as the total protein extract. The protein concentration in each sample was determined using a BCA Protein Assay Kit (Solarbio, Beijing, China) according to the manufacturer’s instructions. Based on the molecular weight of target proteins, equal amounts of protein samples were separated by 10%–12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), then electrotransferred onto nitrocellulose (NC) membranes. The NC membranes were blocked with 5% non-fat dry milk in TBST buffer for 1 h at room temperature to block non-specific binding, followed by incubation with primary antibodies at 4 °C overnight. The primary antibodies used were as follows: Rabbit anti-Claudin 1 (1:1000, 2H10D10, Invitrogen), Rabbit anti-Claudin 5 (1:1000, 4C3C2, Invitrogen), GFAP monoclonal antibody (1:1000, 60190-1-lg, Proteintech), Rabbit anti-tubulin (1:5000, Servicebio) and anti-beta-actin (1:5000, Servicebio). After three washes with TBST, the membranes were incubated with horseradish peroxidase (HRP)-conjugated anti-rabbit IgG secondary antibody (1:5000, Servicebio) for 1 h at room temperature. The protein bands were visualized using an enhanced chemiluminescence (ECL) detection system (Clinx Science Instruments Co., Ltd., Shanghai, China). The relative optical density of each protein band was quantified using Image J software (NIH, USA), with tubulin or beta-actin as the internal reference. Transmission electron microscopy (TEM) Eyes were isolated and fixed in 2.5% glutaraldehyde (pH 7.4) for 2 h. The eye was gently secured in place by the ON and the cornea removed with microdissection scissors rostral or anterior to the limbus. The lens and vitreous were then removed though the opening made by removal of the cornea. After washed three times with 0.1 M phosphate buffer (pH 7.2) and fixed in 1% osmic acid at 4 ℃ for 2 h. Then the samples were gradient dehydrated in a graded series of ethanol. Subsequently, the samples were embedded in Epon-Araldite resin for penetration and placed in a mold for polymerization. After the semi thin section was used for positioning, the ultrathin section was made and collected for microstructure analysis. Followed the counterstaining of 3% uranyl acetate and 2.7% lead citrate. Then observed with a JEM1400 transmission electron microscope. Quantitative real-time PCR analysis Total RNA was extracted from frozen retinal tissues using a Total RNA Extractor Kit (Sangon Biotech, Shanghai, China), and the RNA purity and concentration were determined by a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, USA). First-strand cDNA was synthesized from 1 µg of total RNA using a PrimeScript RT Reagent Kit (Vazyme, Nanjing, China) following the manufacture’s protocol. Quantitative real-time PCR (qRT-PCR) was performed on an ABI 7300 Real-Time PCR System (Applied Biosystems, USA) using a SYBR Green qPCR Master Mix (Vazyme, Nanjing, China). The thermal cycling program was set as follows: initial denaturation at 94 °C for 5 min, followed by 40 cycles of denaturation at 94 °C for 15 s and annealing/extension at 60 °C for 30 s. All qPCR reactions were performed in triplicate with three biological replicates. The primer sequences for target genes are listed in Table 1. The relative mRNA expression levels of target genes were calculated using the 2−ΔCt method, and the fold change in expression compared with the model group was determined using the 2−ΔΔCt method. Tubulin was used as the housekeeping gene for internal normalization. Beta-actin was used as the housekeeping gene for internal normalization. Table 1The mRNA sequences of primersTargetprimerAQP4forward primerCCTCTGCTTTGGACTCAGCAreverse primerCTTTCGTGTGCACACCATGGKir4.1forward primerGGCTATGGCTTCCGCTACATreverse primerAGAATGGTGGTGAGCACGAGBeta-Actinforward primerACCCAGATCATGTTTGAGACCTreverse primerGACCAGAGGCATACAGGGACAACAbbreviations: AQP4, Aquaporin-4; Kir4.1 Inwardly rectifying potassium channel subtype 4.1 Table 1The mRNA sequences of primersTargetprimerAQP4forward primerCCTCTGCTTTGGACTCAGCAreverse primerCTTTCGTGTGCACACCATGGKir4.1forward primerGGCTATGGCTTCCGCTACATreverse primerAGAATGGTGGTGAGCACGAGBeta-Actinforward primerACCCAGATCATGTTTGAGACCTreverse primerGACCAGAGGCATACAGGGACAACAbbreviations: AQP4, Aquaporin-4; Kir4.1 Inwardly rectifying potassium channel subtype 4.1 The mRNA sequences of primers Abbreviations: AQP4, Aquaporin-4; Kir4.1 Inwardly rectifying potassium channel subtype 4.1 Statistical analysis All experimental data were expressed as the mean ± standard error of the mean (SEM). Statistical analyses were performed using SPSS 27.0 software (IBM SPSS, Shanghai, China). The normality of data distribution was tested using the Shapiro-Wilk test. For two-group comparisons, an independent-sample t-test was applied for normally distributed data; otherwise, the Mann-Whitney U test was used. A two-tailed P value < 0.05 was considered statistically significant. Results A decline in visual function in rats following ONTQT To explore the visual function of the ON, we record the F-VEP at 14 days post-modeling (Fig. 2). In terms of P1 peak latency, there was no statistically significant difference between the MG and SG on the 14th day after injury (P > 0.05). However, the peak latency of P2 was prolonged at 14 days post ONTQT injury (P < 0.05). In addition, the MG exhibited a significantly decreased amplitude in N1-P1 and N2-P2 compared to the SG (P < 0.05). The prolonged P2 latency and reduced amplitudes of N1-P1 and N2-P2 in F-VEP indicate that ONTQT surgery decreased visual function of rats, which is consistent with the clinical manifestations of visual dysfunction in patients with ON injury. Fig. 2Evaluation of visual function after modeling measured by F-VEP. A: Representative waveforms of rats in SG. B: Representative waveforms of rats in MG. C: P1 wave peak latency of the two groups (P > 0.05). D: P1 wave amplitude of the two groups (P < 0.05). E: P2 wave peak latency of the two groups (P < 0.05). F: P2 wave amplitude of the two groups (P < 0.05). The data were shown as mean ± SD (n = 6). * P < 0.05 vs. SG Evaluation of visual function after modeling measured by F-VEP. A: Representative waveforms of rats in SG. B: Representative waveforms of rats in MG. C: P1 wave peak latency of the two groups (P > 0.05). D: P1 wave amplitude of the two groups (P < 0.05). E: P2 wave peak latency of the two groups (P < 0.05). F: P2 wave amplitude of the two groups (P < 0.05). The data were shown as mean ± SD (n = 6). * P < 0.05 vs. SG Modifications in the components of NVU within the retina following ONTQT in a rat To observe the structure of the NVU in the inner retina, we conducted TEM on retinal tissue from rats 14 days post-modeling. In the TEM (Fig. 3), we can find the NVU consists of two main components. Firstly, ECs, which are covered by the endothelial basement membrane and sealed tightly by vascular tight junctions, form the BRB. Secondly, the end - feet of astrocytes and Müller glia along with the glial basement membrane make up the glia limitans. These two NVU barriers encircle a perivascular space. The present ultrastructural study showed that microvessels of inner retina after ONTQT exhibited irregularity of luminal circularity compared to that of SG. Remodeling of PCs can be easily observed in the MG sample, which exhibits rounded nuclei and hypertrophic changes accompanied by increased lysosomes, vacuoles, and vesicles. Few Endothelial cell swellings and vacuole formations can be seen in the MG. Furthermore, structural changes of the basement membranes of NVU, including degradation, diffusion and thickening have been found in the retina of MG. Moreover, astrocyte swellings and end-foot detachments were observed in the MG and the perivascular astrocyte showed swollen, cytoplasmic vacuolation and cytoplasmic. Besides, most of the TJs appeared intact as a series of electron-dense zones sealing intercellular clefts in SG. By contrast, ultrastructural abnormalities of TJs were observed in MG. The structural damage to the basement membrane and Tjs observed under TEM reflects the breakdown of the iBRB after ON injury. Fig. 3Transmission electron microscopy image of the retinal NVU, captured in a rat 14-day post modeling. A, C: Representative electron micrograph of the inner retinal NVU in SG and MG. a: red blood cell; b: blood vessel; c: pericytes; d: astrocytes. Scale bar represents 2 μm. B, D: The enlarged detail image of A and C. e: endothelial basement membrane; f: perivascular space; g: glial basement membrane. Scale bar represents 500 nm.SG represented the sham-operated control group. MG represented the model group Transmission electron microscopy image of the retinal NVU, captured in a rat 14-day post modeling. A, C: Representative electron micrograph of the inner retinal NVU in SG and MG. a: red blood cell; b: blood vessel; c: pericytes; d: astrocytes. Scale bar represents 2 μm. B, D: The enlarged detail image of A and C. e: endothelial basement membrane; f: perivascular space; g: glial basement membrane. Scale bar represents 500 nm.SG represented the sham-operated control group. MG represented the model group Loss of RGCs and glia activation in retina of rats after ONTQT The ganglion cell layer (GCL) contained a monolayer of RGCs. The expression of RBPMS, as a marker of RGCs, was used to evaluate the survival of RGC. We used immunofluorescence assays with antibodies against RBPMS in the retina of rats and estimated the expression of RBPMS in GCL 14-day after modeling. As shown in Fig. 4A and B, the number of RGC labeled RBPMS in GCL was reduced in the injured retinal tissues (p < 0.05). To evaluate the activity of glia, GFAP immunofluorescence staining and western blotting were performed. Figure 4C, D and E display GFAP, as a glial marker [16], was over-expression in the injured retinal sections post 14-day ONTQT (p < 0.05). These results illustrate that the modeling of ONTQT significantly promotes the death of RGCs, whereas it enhances the activation of glial cells, which are consistent with pathological features of human ON injury clinically. Fig. 4Retinal immunofluorescent of RBPMS and GFAP. RBPMS immunostaining was applied to label RGC in the GCL of retinal tissues. GFAP immunostaining was used to label reactive gliosis, including astrocytes and Müller glial processes of retinal tissues. A: Representative images of RBPMS stained retinal Sections 14-day post modeling. Green color: RBPMS-positive cells. Blue color: DAPI stained nuclei. B: Quantitative results of RBPMS-positive RGC in the GCL of retinal sections. C: GFAP expression in the retina of rats tested by western blot at 2 weeks post-modeling. D: Representative images of GFAP stained retinal Sections 14-day post modeling. Red color: GFAP-positive protein. Blue color: DAPI stained nuclei. E: Quantitative analysis of GFAP expression on retina sections. The data were shown as mean ± SD (n = 3). * P < 0.05 vs. SG. Scale bar represents 50 μm. SG represented the sham-operated control group. MG represented the model group Retinal immunofluorescent of RBPMS and GFAP. RBPMS immunostaining was applied to label RGC in the GCL of retinal tissues. GFAP immunostaining was used to label reactive gliosis, including astrocytes and Müller glial processes of retinal tissues. A: Representative images of RBPMS stained retinal Sections 14-day post modeling. Green color: RBPMS-positive cells. Blue color: DAPI stained nuclei. B: Quantitative results of RBPMS-positive RGC in the GCL of retinal sections. C: GFAP expression in the retina of rats tested by western blot at 2 weeks post-modeling. D: Representative images of GFAP stained retinal Sections 14-day post modeling. Red color: GFAP-positive protein. Blue color: DAPI stained nuclei. E: Quantitative analysis of GFAP expression on retina sections. The data were shown as mean ± SD (n = 3). * P < 0.05 vs. SG. Scale bar represents 50 μm. SG represented the sham-operated control group. MG represented the model group ONTQT results in a downregulation of tight junction molecules (TJMs) in the retina The maintenance of epithelial-tight junctional integrity by TJMs is essential for protecting the NVU in the retina, where claudins are the primary transmembrane molecules that facilitate epithelial contact. To investigate the expression of TJMs in the retinal NVU, western blot analysis was performed. As shown in Fig. 5A and B, the relative expression of claudin-1 and claudin-5 was reduced in ONTQT retina compared to those in the SG (P < 0.05). Furthermore, astrocytes and Müller cell are integral components of the NVU and play a crucial role in its maintenance. The mRNA expression levels of macroglial end-foot markers, such as AQP4 and Kir4.1, was conducted using qRT-PCR analysis. Figure 5C and D reveal that there was a significant increase in the mRNA level of AQP4 (P < 0.05), whereas the mRNA level of Kir4.1 was decreased in the retina 14 days post-ONC compared to those observed in the SG (P < 0.05). Fig. 5TJM expression in the retina of rats at 2 weeks post-modeling. (A, B) Western blotting was used to assess the relative expression of protein Claudin 1 and Claudin 5 in the retina tissue of rats. Expression was normalized to MG expression level. (C, D) mRNA expression of AQP4 and Kir4.1 in retinal tissue by RT-qPCR detection. Data are presented as mean ± SD (n = 3). * P < 0.05, versus SG. SG represented the sham-operated control group. MG represented the model group TJM expression in the retina of rats at 2 weeks post-modeling. (A, B) Western blotting was used to assess the relative expression of protein Claudin 1 and Claudin 5 in the retina tissue of rats. Expression was normalized to MG expression level. (C, D) mRNA expression of AQP4 and Kir4.1 in retinal tissue by RT-qPCR detection. Data are presented as mean ± SD (n = 3). * P < 0.05, versus SG. SG represented the sham-operated control group. MG represented the model group Discussion In our research, the ONTQT model established recapitulates the key pathological features of TON and the clinical features of TON are characterized by ON damage, visual dysfunction, and retinal nerve fiber layer defect resulted from RGC loss [17]. The disruption of retinal NVU integrity following ONTQT, including TJ breakdown, astrocyte end-foot dysfunction, and glial activation, provides a mechanistic basis for the iBRB impairment and neurovascular uncoupling in TON patients. The retina is a portion of the central nervous system, as well as a highly vascularized tissue [18]. Retinal cells have a high metabolic rate and are extremely sensitive to glucose and oxygen deprivation. The retinal microvasculature supplies nutrients and oxygen to the retinal cells, playing a critical role in maintaining the visual function of the neural retina [19]. Therefore, the microenvironment of the retina must be tightly regulated, and is separated from the systemic circulation by the BRB [20]. The retinal microvasculatures contain the two main cellular constituents of ECs and PCs, sharing the same basement membrane on the blood vessel walls [21]. The basement membrane covered by ECs and PCs constitutes one of the NVU barriers, which tightly sealed by vascular TJs. On the other hand, the glial basement membrane constitutes another barrier within the NVU, comprising the end feet of astrocytes and Müller glia [4]. The NVU plays a role in maintaining the overall integrity of the iBRB, whereas the TJMs located between retinal ECs are the essential molecular structures responsible for the maintenance of iBRB [22]. The TJMs are formed by different types of intracellular proteins and transmembrane adhesion proteins, including the tight-junction-associated MARVEL proteins, claudin family members and junctional adhesion molecules (JAMs) [23]. The claudin family of proteins are transmembrane proteins that allow for the maintenance of tight junction integrity, which have at least 27 known members [24]. Claudins are essential for TJs formation, cell-to-cell adhesion, and the regulation of TJs permeability [1, 25]. Claudin-5 is the most highly expressed claudin protein in endothelial cells of retinal NVU [26]. Claudin-5 expression promotes the sealing of TJs, and as a consequence, decreased vessel permeability, and thus enhanced endothelial barrier function [22]. Diabetic retinopathy (DR) is characterized by increased iBRB permeability. Previous studies have shown that redistribution of claudin-5 and claudin-1 occurs in the diabetic retina, along with increased vascular permeability [27–29]. Inflammation has been implicated in the pathogenesis of DR and inflammatory cytokines like VEGF, TNF-α, IL-6 and IL-1β further decreased claudin-5 expression in the retinas of diabetic animals [27, 30]. Besides, inducible endothelial-specific claudin-5 knockdown mice show severe neuroinflammation within 2 weeks after the initiation of knockdown [31]. Furthermore, previous study has shown that expression of TJM claudin-5 showed a trend towards decreased expression within the first 4 days after ONC [4]. In our study, the expression of claudin-5 and claudin-1 in the retina was significantly downregulated in rats at 14 day following ON injury, indicating the increase permeability of iBRB and the disruption of the NVU in the retina. The downregulation of TJM observed in the study may serve as potential clinical biomarkers for early detection and progression monitoring of ON injury. As an essential component of the visual system, astrocytes perform various functions, including providing nutrients, regulating ionic balance, neurotransmission, supplying the BRB, structural support, and synaptic plasticity [32]. In response to various insults, including glaucoma, ischemia, optic neuritis, ONC, and ON transection, quiescent astrocytes enter a reactive state, the main feature of which is the activation of astrocytes typically exhibiting increased GFAP expression [33]. In retina, astrocytes are located at the inner surface of the retina and Müller glial cells vertically span the entire thickness of the retina, all of which co-regulate the function of NVU. The significant reduction in RBPMS-positive RGCs and upregulation of GFAP expression in the injured retina post ONTQT confirm that glial activation is triggered by ON injury, highlighting glial activation as a potential therapeutic target for neuroprotection in patients with ON damage. As a barrier within the NVU, astrocytic end-feet, which compose the glial basement membrane, envelop the ECs and PCs that constitute the blood vessel lumen in the retina [34]. The study first systematically reveals the multi-dimensional pathological changes in the retinal NVU after ONTQT. Under TEM, ultrastructural abnormalities of NVU at 14th day following ONTQT include irregularity of luminal circularity, PCs remodeling, astrocyte swellings, detachments of astrocyte end-foot, structural changes like degradation, diffusion and thickening of the basement membranes of NVU, all of which were consistent with previous studies in CNS injuries model [35, 36]. However, swollen ECs were not distinct in the retina of NVU at 14th day following ONTQT, which may relate to the observe time. Our findings provide a direct mechanistic link between ON injury, NVU disruption, and visual loss, offering a pathological basis for clinical observations of iBRB dysfunction in optic neuropathy patients. In the eyes, AQP4 is found to be expressed in the retinal astrocytes and Müller glial cells, with the highest expression occurring in the end feet membranes, which are major in transcellular water transport and in regulating cellular volume [37]. In the mammalian retina, Kir4.1 is mostly expressed in the end feet of retinal Müller glial cells and is the main potassium channel in Müller cells, whose polarised expression allows K+influx into Müller cells from the extracellular space [38]. Water movement into and out of Müller cells relies on the co-expression of Kir4.1 and AQP4 on Müller cells, which allows water to follow K+ from perisynaptic spaces to blood vessels [39]. AQP4 and Kir4.1 in retinal Müller cells play important roles in controlling retinal potassium, water homeostasis, as well as maintaining the integrality of iBRB [40]. In diabetic rat both BRB disruption and retinal oedema were also observed in these retinas as well as the increased AQP4 levels and decreased Kir4.1 levels [41]. In retinal injury induced by hypobaric hypoxia, the breakdown of the iBRB leads to retinal edema and increased expression of GFAP in Müller cells, accompanied by an increase in retinal AQP4 expression and a decrease in Kir4.1 expression [39]. Our results showed the downregulation of Kir4.1 mRNA levels in the retina, which coincide with the previous studies on different models of ON and retina injury [39, 41, 42]. The decline mRNA expressions of Kir4.1 in our research indicates the intracellular potassium accumulation, which may lead to increased intracellular and extracellular osmotic pressure in astrocytes. Moreover, AQP4 mRNA levels in retina of rats decreased at days 2, 7 and 14 following ONC [42]. However, the previous research found augmented AQP4 mRNA level in the retina of mice at 2 and 3 days post-ONC [4, 43]. In our research, mRNA expressions of AQP4 in retina increased at days 14 following ONTQT injury. Therefore, AQP4 mRNA level of retina vary depending on the different damages, species and the stage of injury. The increasing number of reactive astrocytes in the inner retina following ONTQT may account for the rising mRNA level of AQP4. Although we did not assess AQP4 protein in this study, the upregulation of AQP4 mRNA may reflect altered function of astrocyte, which could potentially be associated with the development of glial edema observed under TEM. However, it should be noted that mRNA expression does not necessarily correlate with protein function and distribution. further studies including immunofluorescence and western blot analysis of AQP4 need to be performed to confirm these findings. Retinal pathologies are accompanied by alterations of amounts and/or spatial distribution of AQP4 or Kir4.1 [38]. Müller cell oedema due to the regulation or redistribution of Kir4.1 and AQP4 has been reported to be involved in the development and progression of a variety of retinal diseases, including diabetic macular oedema, retinal vein occlusion and uveitis [39]. The dysregulation of macroglial end-foot markers AQP4 and Kir4.1 following ON injury in the research, provides critical insights into the molecular mechanisms of NVU dysfunction in patients with TON. Firstly, the ON traction stimulates an acute local response in the injury epicenter and exacerbated glial activation immediately [44] so that increased GFAP expression of retina was exhibited following ONTQT. In the research, the reactive astrocyte with the abnormal mRNA expression of AQP4 and Kir4.1 cannot balance the potassium and water homeostasis resulting in the astrocyte swellings and detachments of astrocyte end-feet. Furthermore, another barrier of NVU was impaired, which was the basement membrane covered by ECs and PCs and tightly sealed by vascular TJs. The expression of TJs proteins like claundin-1 and claudin-5 decreased and under TEM the degradation, diffusion and thickening of the basement membranes of NVU can be observed. Finally, the disfunction of NVU boost the RGC loss in the inner retina, which was exhibited by the decrease number of RBPMS-positive RGC following ONTQT. Meanwhile, the RGC loss resulted in the visual disfunction tested by FVEP at the 14th day following ONTQT. Conclusions In conclusion, ONTQT as a new model of indirect TON can reliably induce ON injuries and RGC loss. The research reveals that there is a link between the dysfunction of NVU and the death of RGC at 14th day following ON injury. The mechanisms of iBRB breakdown after ONTQT are relative to the downregulation of the TJs like claudin-1 and claudin-5 and maybe associated with the altered mRNA expression of perivascular macroglial end feet markers like AQP4 and Kir4.1. The two barriers of NVU, constituted by the ECs and PCs and the end feet of astrocytes and Müller glia, are critical for RGCs protection after ON injury. The study highlights the retinal NVU as a novel therapeutic target for neuroprotection in ON injury. Additionally, targeting these molecules including stabilize TJs and restore Kir4.1 function, represents a promising translational strategy for developing novel neuroprotective therapies to preserve RGC survival and visual function in patients with TON. However, we are lack of experiments at different time points after ONTQT. The expression of TJ proteins and perivascular macroglial end feet markers may show varying outcomes in the early stages following ON injury. The further investigation should be developed in the future. Electronic Supplementary Material Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2