Guiding principles of the Declaration of Helsinki were strictly followed in all present procedures. All human participants were recruited from the Department of Ophthalmology, The First Affiliated Hospital of Chongqing Medical University (Chongqing, China). Serum samples were collected between March 2024 and October 2025. The entire study protocol was approved by the Ethics Committee of the First Affiliated Hospital of Chongqing Medical University (Chongqing, China; approval no. 2024-087-01). Written informed consent was obtained from all participants involved in the research.

To investigate the differences in ketone body levels between open-angle and angle-closure glaucoma, 15 patients with primary open-angle glaucoma (POAG), 15 patients with primary angle-closure glaucoma (PACG) and 17 age- and sex-matched healthy controls were initially enrolled and analyzed. To further evaluate the effect of IOP on ketone body levels specifically within POAG, an additional 15 POAG patients were enrolled, expanding the total POAG cohort to 30 individuals. These 30 patients were subsequently stratified by baseline IOP into a high-IOP POAG subgroup (POAG-H; IOP >21 mmHg; n=15) and a normal-IOP POAG subgroup (POAG-N; IOP ≤21 mmHg; n=15) (24). For the subgroup comparison, 15 of the 17 original controls were selected through 1:1 matching based on age and sex to serve as the matched reference group. The two unmatched controls were excluded from the subgroup analyses but remained in the initial three-group comparison. Visual field defects were assessed using a Humphrey Field Analyzer (Carl Zeiss AG) with the 24-2 Swedish Interactive Thresholding Algorithm standard strategy (25). Peripheral blood samples were collected from each participant and processed to isolate serum, which was stored at -80°C. Patients with a clinical diagnosis of glaucoma (including both POAG and PACG), confirmed by at least two experienced ophthalmologists, aged ≥18 years and capable of providing written informed consent, were enrolled in this study. Patients with glaucoma of any disease duration, regardless of prior treatment status, were included. Individuals with other ocular diseases (such as diabetic retinopathy and age-related macular degeneration), hypertension, diabetes mellitus, systemic infections or any condition that could affect systemic metabolism were excluded. The present study enrolled 62 participants (32 men and 30 women), with a median age of 65 (range: 50-81 years). The details for each group, including the numbers sex distribution, median age and age range, are summarized in Table SI. Detailed clinical characteristics and visual field parameters for the POAG-H and POAG-N subgroups are presented in Table SII.

All animal procedures were approved by the Institutional Animal Care and Use Committee of Chongqing Medical University (Chongqing, China; approval no. IACUC-CQMU-2025-0349) and performed in compliance with the Association for Research in Vision and Ophthalmology guidelines (26). A total of 72 adult male C57BL/6J mice (6-8 weeks old) were used in the present study, with 15 being used for the pre-experiment. Mice were divided into a control group (CON; n=3) and an EAG group (12 mice, subdivided into four time points at weeks 2, 4, 6 and 8; n=3 each). Due to the limited amount of retinal tissue obtained from each mouse, to ensure a sufficient sample size for multiple assays (n=3-6 per group), 57 mice were used for the formal experiments, including the CON group (6 mice), the EAG group (6 mice), the CON + control diet group (15 mice), the EAG + control diet group (15 mice) and the EAG + KD group (15 mice). The exact n values for each specific assay are indicated in the corresponding figure legends.

Mice were obtained from the Experimental Animal Center of Chongqing Medical University (Chongqing, China) and housed in a pathogen-free environment. EAG was induced as previously described (27,28). Briefly, fresh bovine eyes were obtained from a local slaughterhouse (Chongqing, China) for optic nerve antigen (ONA) extraction, ground at -30°C and then suspended in PBS (cat. no. BL601A; Biosharp Life Sciences) to obtain a suspension with a concentration of 1 mg/ml. For immunization, the mice received an intraperitoneal injection of 50 µl ONA emulsified with the same volume of incomplete Freund's adjuvant (FA) (cat. no. P2031; Beyotime Biotechnology), supplemented with 1 µg pertussis toxin (PTX; cat. no. P7208; Sigma-Aldrich; Merck KGaA). A total of 2 days later, these mice received another injection of PTX (1 µg) through the same route. PTX was administered to ensure the permeability of the blood-retinal barrier as previously described (27). The control mice received intraperitoneal injection of 100 µl PBS, FA and PTX.

For the pre-experiment, mice were euthanized at weeks 2, 4, 6 and 8 post-immunization to monitor disease progression and determine the optimal endpoint for therapeutic evaluation. For formal experiments, all mice were euthanized at week 4 post-immunization. At each designated time point, serum samples (~100 µl) and retinal tissues were harvested for subsequent analysis. For blood collection, mice were deeply anesthetized with tribromoethanol (250 mg/kg; cat. no. 2092A; Nanjing Aibei Biological Technology Co., Ltd.) administered through intraperitoneal injection. Following the induction of deep anesthesia (determined by the loss of pedal withdrawal reflex), blood was collected by enucleation of the eyeball. Immediately following blood collection, mice were subsequently euthanized through cervical dislocation. Mortality was determined by the complete cessation of heartbeat and respiration, as well as the absence of all reflexes. Retinal tissues were then harvested immediately after euthanasia. Serum was separated from whole blood samples by centrifugation (100 × g for 20 min at room temperature) and stored at -80°C until further analysis. Blood samples and retinal tissues were collected from separate cohorts of mice at each time point and all mice were euthanized immediately after each sample collection.

Semi-purified diets, including a KD (cat. no. D10070801) and a control diet (CD; cat. no. D10070802), were purchased from Ready Dietech (Shenzhen) Co., Ltd. Their detailed compositions are provided in Table SIII. Immunized mice were randomly assigned to two groups and fed ad libitum for 4 weeks with either an irradiated KD (EAG-KD) or an irradiated CD (EAG-CD). A separate cohort of non-immunized mice was fed the CD (CON-CD) to serve as a negative control. Upon termination of the present study, retinal tissues and serum samples were collected from all mice for further experimental procedures.

All procedures were supported by the Institutional Animal Care and Use of Chongqing Medical University (Chongqing, China; approval no. IACUC-CQMU-2025-0349) and in compliance with Association for Research in Vision and Ophthalmology guidelines. The culture procedures for primary Müller cells followed a previously published report with minor modifications (29). Briefly, Müller cells were isolated from retinas of 5-day-old C57BL/6J mice (mixed-sex neonatal mice). Based on the present preliminary optimization experiments, retinas from 8 neonatal mice were pooled and seeded into a single T75 culture flask, which consistently yielded sufficient cell numbers with high purity after passaging. To ensure experimental reproducibility, four independent isolations were performed for the present study, using a total of 32 neonatal mice.

Neonatal mice were euthanized by cervical dislocation before tissue collection. Mortality was determined by complete cessation of heartbeat and respiration, as well as absence of reflexes. Following dissection, retinal pieces were digested in 0.25% trypsin (cat. no. 25200072; Gibco; Thermo Fisher Scientific, Inc.) for 3-5 min at 37°C to obtain separate cells. Dissociated cells were seeded into poly-d-lysine (cat. no. P2100; Beijing Solarbio Science & Technology Co., Ltd.) pre-coated 75 cm² flasks, and then maintained in DMEM (cat. no. 11965092; Gibco; Thermo Fisher Scientific, Inc.) containing 20% FBS (cat. no. NEW500; Serana Biologicals) and 1% penicillin-streptomycin (cat. no. C0222; Beyotime Biotechnology) under standard conditions (37°C and 5% CO₂). Medium changes were performed at intervals of 3-4 days. The culture medium was replaced twice a week. Subsequently, ~10 days later, the cells were passaged at a 1:1 ratio. After 2-3 passages at the same split ratio (1:1), the Müller cells became relatively pure and were identified by cell morphology and immunofluorescence with glutamine synthetase (GS). Müller cells in passages 3-5 were used for the present experiments.

For lipopolysaccharide (LPS; cat. no. L2880; Sigma-Aldrich; Merck KGaA) treatment, Müller cells, seeded in DMEM with 5% FBS, were then subjected to LPS (0, 0.1, 0.5, 1, 2, 5 or 10 µg/ml) for 72 h at 37°C. For BHB (MedChemExpress; cat. no. HY-113378) treatment, Müller cells maintained in DMEM with 5% FBS were exposed to a range of BHB concentrations (0, 0.5, 1, 2, 5, 10, 20 or 50 mM) for 72 h at 37°C, with or without co-treatment of 1 µg/ml LPS.

Serum was separated from the peripheral blood of human participants and mice, followed by the measurement of serum concentrations of BHB and acetoacetic acid using a microplate assay kit (cat. no. BC5065; Beijing Solarbio Science & Technology Co., Ltd.) according to the manufacturer's instructions. Briefly, the working solutions of BHB and AcAc were separately mixed with the serum and placed in a 96-well microplate for reaction. The optical density (OD) at 340 nm was measured using a Thermo Scientific Varioskan™ LUX microplate reader (Thermo Fisher Scientific, Inc.) at 37°C and then statistical calculations were conducted.

Serum was separated from mouse peripheral blood, followed by the measurement of serum glucose concentrations using a colorimetric assay kit (cat. no. G1212W; Suzhou Grace Biotechnology Co., Ltd.) according to the manufacturer's instructions. Briefly, the working solution was mixed with the serum and placed in a 96-well plate at 37°C in the dark for 30 min. The OD at 510 nm was measured at 37°C using a Thermo Scientific Varioskan LUX microplate reader (Thermo Fisher Scientific, Inc.) and then statistical calculations were conducted.

IOP was measured weekly from 9:00-11:00 AM using a TonoLab rebound tonometer (iCare Finland Oy) until the experimental endpoint. Each recorded value constitutes the mean of six independent IOP measurements, as described in a previous study by the present authors (30).

For OCT imaging, general anaesthesia was induced to mice through intraperitoneal injection of tribromoethanol (250 mg/kg). Following anaesthesia, pupils were dilated using compound tropicamide eye drops to enable clear fundus visualization. Mice were subjected to OCT using an Ultramicro Ophthalmol Imaging System (ISOCT; Optoprobe Science Ltd.) for assessment of the ganglion cell layer (GCL) thickness. All OCT scans, performed using the Optoprobe-OCT system, were centred on the posterior pole of the retina. To ensure comparability, scans were consistently acquired from the same retinal location. The system simultaneously captured fundus images during scanning and generated detailed fundus maps. Subsequently, the acquired OCT images were analyzed using the OCT Image Analysis Software (version 2.0; Optoprobe Science Ltd.) to obtain the average GCL thickness. After OCT acquisition, mice were allowed to recover from anaesthesia.

Mouse eyeballs were harvested and immersion-fixed in undiluted FAS Eyeball Fixative Solution (cat. no. G1109; Wuhan Servicebio Technology Co., Ltd.) at room temperature for at least 24 h before paraffin embedding. Consecutive 4-µm sections along the corneal-optic nerve axis were generated. H&E staining was performed at room temperature using the H&E Staining Kit (cat. no. G1076; Wuhan Servicebio Technology Co., Ltd.) according to the manufacturer's standard operating procedures. Briefly, sections were dewaxed in dewaxing solution (cat. no. G1128; Wuhan Servicebio Technology Co., Ltd.) for 15 min twice, rehydrated in graded ethanol (100% ethanol twice, 75% ethanol once, 5 min each; cat. no. 100092683; Sinopharm Chemical Reagent Co., Ltd.), and washed with tap water. Sections were then pretreated with High-Definition Constant Temperature Pretreatment Solution (Servicebio) for 1 min, followed by staining with hematoxylin for 3 min. After washing, sections were differentiated in differentiation solution (Wuhan Servicebio Technology Co., Ltd.) for 5 sec, blued in bluing solution (Wuhan Servicebio Technology Co., Ltd.) for 5 sec and rinsed with tap water. Subsequently, sections were dehydrated in 95% ethanol for 1 min and stained with eosin for 15 sec. Finally, sections were dehydrated through graded ethanol (100% ethanol three times, 1 min each) and n-butanol (cat. no. 100052190; Sinopharm Chemical Reagent Co., Ltd.) twice for 1 min each, cleared in xylene (cat. no. X100301-S5L; Shanghai Lingfeng Chemical Reagent Co., Ltd.) twice for 1 min each, and mounted with neutral balsam (cat. no. 10004160; Sinopharm Chemical Reagent Co., Ltd.) Stained sections were imaged using a conventional light microscope (Leica Microsystems GmbH).

Following enucleation, eyeballs were fixed by immersion in FAS Eyeball Fixative Solution for 2 h at room temperature. The retinas were dissected and flat-mounted and then permeabilized with 0.5% Triton X-100 (cat. no. P0096; Beyotime Biotechnology) for 30 min and blocked in a solution containing 5% normal goat serum (cat. no. AR0009; Boster Biological Technology, Co., Ltd.) at room temperature for 1 h. Subsequently, retinas were incubated overnight at 4°C with the RGC specific marker brain-specific homeobox/POU domain protein 3A (Brn3a; 1:100; cat. no. ab81213; Abcam) and then incubated with a 594-conjugated goat anti-rabbit IgG (H+L) secondary antibody (1:200; cat. no. AS039; ABclonal Biotech Co., Ltd.) at room temperature for 1 h and counterstained with DAPI at room temperature for 5 min (cat. no. P0131; Beyotime Biotechnology) to visualize nuclei. Images were captured on an immunofluorescence microscope (Leica Microsystems, Inc.).

Deparaffinized sections (cat. no. G1128; Servicebio) underwent heat-induced epitope retrieval in Tris-EDTA buffer (pH 9.0; 95°C; 30 min). After blocking with 5% normal goat serum at room temperature for 1 h, sections were probed with primary antibodies against glial fibrillary acidic protein (GFAP; 1:300; cat. no. 60190-1-Ig), monocarboxylate transporter (MCT)-1 (1:200; cat. no. 20139-1-AP), MCT2 (1:300; cat. no. 20355-1-AP) and 3-oxoacid CoA-transferase 1 (OXCT1; 1:200; cat. no. 12175-1-AP; all, Proteintech Group, Inc.) overnight at 4°C. Detection was achieved using Alexa Fluor 488-goat anti-rabbit (cat. no. A0423) and Alexa Fluor 555-donkey anti-mouse (cat. no. A0460) secondary antibodies (all, 1:500; all, Beyotime Biotechnology) for 1 h at room temperature, followed by DAPI counterstaining at room temperature for 5 min. Images were captured on an immunofluorescence microscope (Leica Microsystems, Inc.).

For the immunofluorescence staining of Müller cells, briefly, cells from passages 2-3 were seeded on glass slides in a 24-well plate for 24 h. After fixation with 4% paraformaldehyde (cat. no. P0099; Beyotime Biotechnology) at 37°C for 10 min and permeabilization with 0.5% Triton X-100 (cat. no. P0096; Beyotime Biotechnology) at 37°C for 30 min, samples were blocked with 5% BSA (cat. no. FA016; Genview) at 37°C for 30 min and incubated overnight at 4°C with primary antibody against GS (1:200; cat. no. GTX109121; GeneTex, Inc.). Subsequently, cells were incubated with Alexa Fluor 488-conjugated goat anti-rabbit IgG (H+L) (as aforementioned) for 1 h at room temperature. Slides were mounted in anti-fade medium containing DAPI and visualized using an immunofluorescence microscope (Leica Microsystems, Inc.).

A Cell Counting Kit-8 (CCK-8; cat. no. HY-K0301; MedChemExpress) was employed to determine cell viability. Following the manufacturer's instructions, 100 µl cells (1×10⁶ cells/ml) were inoculated into a 96-well plate. The next day, cells were cultured with numerous concentrations (0.1, 0.5, 1, 2, 5 or 10 µg/ml) of LPS at 37°C for 72 h or treated with various concentrations (0, 0.5, 1, 2, 5, 10, 20 or 50 mM) of BHB with or without 1 µg/ml LPS at 37°C for 72 h. Following a 2-h incubation with 10 µl CCK-8 solution at 37°C in the dark, the OD at 450 nm of Müller cell cultures was quantified using a Varioskan LUX microplate reader (Thermo Fisher Scientific, Inc.). All assays were performed at least three times.

Proteomics analysis was performed by Shanghai Applied Protein Technology Co., Ltd. A total of three independent cultures of Müller cells (2×10⁶ cells per 10 cm dish) were prepared for each group (Control, LPS and LPS + BHB). After 72 h of treatment at 37°C, the culture medium was removed and cells were washed, scraped and collected into 1.5 ml tubes. After sample collection, samples were lysed in SDC buffer (5% sodium deoxycholate; 100 mM Tris-HCl; pH 8.5), followed by protein reduction and alkylation using 1 mM Tris (2-carboxyethyl) phosphine and 1 mM chloroacetamide, respectively. After spiking with indexed retention time peptides, the samples were analysed using an Evosep One-coupled timsTOF Pro2 mass spectrometer (Bruker Daltonics; Bruker Corporation) in DIA mode. The resulting 4D-DIA data were processed using Spectronaut™ (version 19.7) software (31,32). Proteins exhibiting decreased expression in the LPS group relative to control and increased expression in the LPS + BHB group relative to LPS were first identified. From these, proteins meeting the criteria of fold change >1.2 and P<0.05 in pairwise comparisons were selected for further analysis. The present mass spectrometry proteomics data were deposited in the iProX database (https://www.iprox.cn) under accession no. IPX0018867000.

Total RNA was extracted from mouse retinal tissues and cultured Müller cells using Total RNA Isolation Reagent (cat. no. AG21017; Accurate Biotechnology (Hunan) Co., Ltd.). For cDNA synthesis, 500 ng of RNA was first treated with 5X gDNA digester Mix (cat. no. HY-K0511A; MedChemExpress) at 42°C for 2 min in a 15-µl reaction to remove genomic DNA. Subsequently, 4X Super RT Mix (from the same kit) was added to a final volume of 20 µl, and reverse transcription was performed at 25°C for 5 min, 55°C for 15 min and 85°C for 2 min. The resulting cDNA was stored at 4°C until use for qPCR. Gene expression was quantified by qPCR on an Applied Biosystems 7500 system using SYBR Green qPCR Master Mix (cat. no. HY-K0522; MedChemExpress). The qPCR cycling conditions were as follows: Initial denaturation at 94°C for 10 sec; 4 cycles of 50°C for 30 sec and 72°C for 1 min; followed by 95°C for 30 sec; and 42 cycles of 95°C for 5 sec, 60°C for 30 sec and 72°C for 30 sec. Transcript levels for GFAP, IL-1β, IL-6, TNF-α, MCT1, MCT2 and OXCT1 were calculated using the 2^−ΔΔCq method relative to β-actin (33). Primer sequences are listed in Table SIV.

Retina and Müller cell proteins were extracted using a lysis buffer composed of RIPA Lysis Buffer (cat. no. P0013B; Beyotime Biotechnology) and PMSF solution (cat. no. ST506; Beyotime Biotechnology) at a ratio of 99:1. Protein concentrations were quantified using a BCA method (cat. no. P0009; Beyotime Biotechnology). Equal amounts of protein (20 µg per lane) were resolved on 4-20% gels using SDS-PAGE (cat. no. F15412MGel; ACE Biotechnology) and transferred to PVDF membranes (cat. no. IPVH00010; MilliporeSigma; Merck KGaA). Blocking was performed using QuickBlock™ Buffer (cat. no. P30500; NCM Biotech) for 20 min at room temperature. Membranes were probed overnight at 4°C with specific primary antibodies, followed by incubation with secondary antibodies HRP-conjugated Goat Anti-Rabbit IgG(H+L) (1:10,000; cat. no. SA00001-2; Proteintech Group, Inc.) and Dylight 800 Goat Anti-Mouse IgG (1:20,000; cat. no. RS23910; ImmunoWay Biotechnology Company) for 1 h at room temperature. Protein bands were visualized using a Western Bright™ ECL kit (cat. no. HY-K2005; MedChemExpress). The relative protein expression levels were quantified by grayscale analysis using the ImageJ software (version 1.54p; National Institutes of Health). Specifically, the intensity of the acetylated FOXO3A band was normalized to the intensity of the corresponding total FOXO3A band and then further normalized with β-tubulin as the loading control. All steps were performed according to the manufacturers' protocols.

Primary antibodies used included: GFAP (1:3,000; cat. no. 60190-1-Ig; Proteintech Group, Inc.), MCT1 (1:1,000; cat. no. 20139-1-AP; Proteintech Group, Inc.), MCT2 (1:1,000; cat. no. 20355-1-AP; Proteintech Group, Inc.), OXCT1 (1:1,000; cat. no. 12175-1-AP; Proteintech Group, Inc.), IL-1β (1:1,000; cat. no. WL02257; Wanleibio Co., Ltd.), IL-6 (1:1,000; cat. no. WL02841; Wanleibio Co., Ltd.), TNF-α (1:1,000; cat. no. WL01581; Wanleibio Co., Ltd.), acetylated-lysine (1:2,000; cat. no. HA723073; HUABIO), FOXO3A (1:2,000; cat. no. ET1604-11; HUABIO), acetyl-FOXO3A (Lys271; 1:2,000; cat. no. AF3771; Affinity Biosciences), acetyl-FOXO3A (Lys290; 1:1,000; cat. no. AF3770; Affinity Biosciences), MT2A (1:2,000; cat. no. DF6755; Affinity Biosciences), β-actin (1:20,000; cat. no. HA722023; HUABIO) and β-tubulin (1:20,000; cat. no. ET1602-4; HUABIO).

All data are presented as the mean ± SD. Statistical analyses were performed using SPSS (version 26.0; IBM, Corp.). GraphPad (version 10.0; Dotmatics) was used to visualize all experimental data. Normality was assessed using the Shapiro-Wilk test. Depending on normality, two-group comparisons were made using unpaired t-tests or Mann-Whitney U tests and multi-group comparisons used one-way ANOVA with Tukey's multiple comparisons test or the Kruskal-Wallis tests with Dunn's post-hoc test. The n values represent the number of samples. A minimum of three independent replicates were included for all experiments. Statistical significance is denoted as follows: ^*P<0.05, ^**P<0.01, ^***P<0.001 and ^****P<0.0001. Results

Serum levels of BHB and AcAc were first measured in patients with POAG, patients with PACG and healthy controls to identify differences among the three groups (Fig. 1A). The POAG group exhibited significantly lower serum ketone body levels compared with the PACG and healthy control groups (Fig. 1B and C). To determine whether decreased serum ketone body levels were associated with IOP in patients with POAG, patients were divided into a POAG-H group (30.87±9.70 mmHg) and a POAG-N group (17.67±1.95 mmHg), based on whether their first IOP measurement on admission was >21 mmHg (Fig. 1D). Levels of BHB and AcAc were significantly reduced in both the POAG-H and POAG-N groups relative to controls (Fig. 1E and F). Baseline characteristics of the POAG-H and POAG-N groups, including disease duration, medication use and visual field defects, are summarised in Table SII. No significant differences in these parameters were observed between the two subgroups, suggesting that the reduction in ketone body levels was independent of IOP and other clinical confounders.

Subsequently, an ONA-induced EAG model was employed (Fig. 2A). Quantification of RGCs by immunofluorescence at 2, 4, 6 and 8 weeks after immunization revealed statistically significant RGC loss beginning at week 4 (Fig. S1A). RT-qPCR analysis of IL-1β, IL-6, TNF-α and GFAP at 0, 2 and 4 weeks exhibited progressive upregulation of these markers, with a significant increase at week 4 (Fig. S1B). Collectively, these results demonstrated successful establishment of the EAG model by week 4 and supported its use in subsequent experiments. During the first 4 weeks after immunization, the IOP remained normal in the EAG and the control groups, with no significant differences observed (Fig. 2B). H&E staining and OCT images exhibited significant thinning of the GCL and loss of RGCs in the EAG group compared with controls (Fig. 2C and D). RGC loss was further determined by Brn3a immunostaining, which showed a significant reduction in RGC numbers in the EAG group compared with the control group (Fig. 2E and F). The association between ketone body levels and glaucomatous damage was then examined, revealing significantly lower serum ketone body levels in the EAG group compared with the control group, consistent with observations in POAG (Fig. 2G). These results suggest a possible association between optic nerve injury and reduced serum ketone body levels in EAG mice.

KD feeding was used to increase ketone body levels in vivo. EAG mice were fed a CD (80% CHO, 10% protein and 10% fat) or a KD (0% CHO, 10% protein and 90% fat) for 4 weeks (Fig. 3A). Serum ketone body levels were found to be significantly elevated in KD-fed EAG mice compared with those in the CD-fed group (Fig. 3B). ONA immunization did not affect blood glucose levels but resulted in body weight loss in mice. By contrast, KD had no significant effect on either blood glucose levels or body weight in EAG mice (Fig. 3C and D).

To evaluate changes in ketone body transport and utilization in the retina, key enzymes and molecules involved in these processes were examined. The expression of monocarboxylate transporters (MCT1 and MCT2) and OXCT1 was found to be significantly decreased in the retinas of EAG mice compared with healthy controls at both the protein and RNA levels. Following KD intervention, MCT1 and OXCT1 expression increased, whereas MCT2 expression exhibited no significant change (Fig. 3E-I). These results demonstrated that KD increased serum ketone body levels and concurrently upregulated retinal MCT1 and OXCT1 expression in EAG mice.

To evaluate the effects of elevated serum ketone bodies on EAG mice, retinal layer thickness was assessed by H&E staining, with OCT and RGCs being quantified by immunofluorescence. KD-fed EAG mice exhibited increased GCL thickness and significantly elevated RGC numbers compared with the CD-fed EAG group (Fig. 4A-D). A hallmark of Müller cell activation is the upregulation of GFAP (10). Immunofluorescence, western blotting and RT-qPCR analyses revealed significantly increased GFAP expression in the retina of EAG mice compared with the control, which was further significantly reversed by KD feeding (Fig. 4E-H). Similarly, western blotting and RT-qPCR analyses showed significantly elevated expression of IL-1β, IL-6 and TNF-α in the retina of EAG mice, with KD feeding further significantly reducing their levels (Fig. 4I and J). These findings indicate that KD was associated with suppressed Müller cell activation, reduced inflammatory factor expression and improved RGC survival in the EAG retina.

To explore the role of BHB in vitro, primary Müller cells were isolated from neonatal mouse retinas and cultured. After 2-3 passages, the cells achieved high purity, as determined by their distinct morphology and positive immunofluorescence for GS (Fig. 5A; Fig. S2). Müller cell viability after 72 h of LPS stimulation was measured using a CCK-8 assay. Cell viability decreased significantly at all tested LPS concentrations >1 µg/ml (2, 5 and 10 µg/ml). Based on this, 1 µg/ml was chosen as the maximum concentration that did not compromise cell viability and was used for subsequent experiments (Fig. 5B). The effects of BHB on Müller cells were then evaluated with or without 1 µg/ml LPS. In the absence of LPS, BHB concentration up to be 20 mM did not significantly affect cell viability (Fig. 5C). Under LPS stimulation, no significant reduction in cell viability was observed at BHB concentrations up to 5 mM, whereas a significant decline occurred at 10 mM (Fig. 5D). Accordingly, 5 mM BHB was selected as the maximum non-toxic concentration for further experiments. After 72 h of LPS stimulation, MCT1 and OXCT1 expression was found to be significantly reduced in Müller cells compared with controls. In the LPS-stimulated cells treated with BHB, MCT1 and OXCT1 expression was significantly upregulated (Fig. 5E and F). Concurrently, LPS stimulation significantly increased GFAP expression in Müller cells and further significantly reduced it upon BHB treatment (Fig. 5G and H). Western blotting and RT-qPCR analyses showed that LPS-induced increases in IL-1β, IL-6 and TNF-α were significantly suppressed by BHB treatment (Fig. 5I and J). Collectively, these in vitro findings indicate that BHB is associated with reduced Müller cell activation and inflammatory factor secretion, consistent with the in vivo observations.

KD exerts a similar inhibitory effect to a naturally occurring and specific inhibitor of class I histone deacetylases. Pan-acetylation was assessed by detecting overall acetylated lysine levels. KD feeding and BHB treatment restored the diminished global acetylation levels in the retinas of EAG mice and in LPS-stimulated Müller cells, respectively (Fig. 6A). Given the observed global increase in protein acetylation, 4D-DIA proteomic analysis was performed to pinpoint the specific proteins and pathways affected. Comparative analysis revealed distinct protein expression profiles among the control, LPS and LPS+BHB groups (Fig. 6B and C; Fig. S3A-D). A set of 76 proteins that were decreased in the LPS group relative to controls and increased in the LPS + BHB group was identified and visualized using a clustering heatmap (Fig. 6D; Table SV). Among these, proteins that met the criterion of fold change >1.2 and had a P<0.05 in paired comparisons were selected. FOXO3A and MT2A were found to be significantly upregulated and their expression differences are presented as a volcano plot and a box plot, respectively (Fig. 6E-G). GO and KEGG enrichment analyses revealed significant upregulation of the 'FoxO signalling pathway' following BHB treatment in LPS-stimulated cells. (Figs. 6H and S3E). Reactome analysis further supported the involvement of FOXO, with the most prominent enrichment observed in 'Regulation of FOXO transcriptional activity by acetylation' (Fig. 6I).

Western blotting analysis demonstrated that LPS significantly suppressed FOXO3A and MT2A expression, as well as FOXO3A acetylation at Lys271, in Müller cells and these effects were then significantly reversed by BHB treatment (Fig. 7A). Consistent with these in vitro findings, parallel results were observed in the retina of EAG mice following KD feeding (Fig. 7B). Acetylation at Lys290, another known acetylation site of FOXO3A, was also investigated. In retinal tissue, KD feeding significantly increased FOXO3A acetylation at Lys290, a change similar to the trend observed at Lys271. However, in cultured Müller cells, Lys290 acetylation was barely detectable and did not significantly change upon BHB treatment (Fig. S4). These results indicate that the protective effect of BHB on Müller cells was accompanied by increased FOXO3A acetylation at Lys271 and upregulation of MT2A.