Resveratrol attenuates diabetic RGC degeneration through TRIM23-dependent VDAC1 ubiquitination and mitophagy

Abstract

Background

Retinal ganglion cells (RGCs) degeneration is an early event in diabetic retinopathy (DR), tightly coupled to mitochondrial dysfunction. While the mitochondrial gatekeeper voltage-dependent anion channel 1(VDAC1) is a known mediator of apoptosis, its regulation and therapeutic targeting in diabetic RGCs remain unclear.

Methods

We examined VDAC1 expression in human donor retinas, streptozotocin-induced diabetic mice, and high glucose-treated SH-SY5Y cells. VDAC1-knockdown SH-SY5Y cells were established to explore its functional role in the context. Using virtual screening and molecular docking, we identified resveratrol (RSV) as a potential VDAC1 inhibitor, which was further validated by surface plasmon resonance binding assays and RSV-based pull-down experiments. The effects of RSV were assessed by RGCs survival, retinal electrophysiological function, and mitochondrial quality control. Tripartite motif-containing 23 (TRIM23)-VDAC1 interaction along with ubiquitination modifications were confirmed by co-immunoprecipitation. Finally, combination therapy involving RSV and adeno-associated virus serotype 2 (AAV2) carrying the γ-synuclein (SNCG) promoter mediated Trim23 overexpression was applied in both DR and acute ocular hypertension (AOH) animal models.

Results

VDAC1 was upregulated in diabetic RGCs, and its knockdown was protective. We further discovered that RSV recruits the E3 ligase TRIM23 to VDAC1, inducing a K27-linked ubiquitination switch that promotes mitophagy instead of apoptosis. We propose that RSV binding acts as an allosteric switch, reconfiguring VDAC1 to favor this protective ubiquitination pathway. Consequently, RSV restored mitochondrial health and RGCs viability. Therapeutically, enhancing this axis via RSV and RGCs-targeted Trim23 overexpression synergistically protected against RGCs loss in both DR and AOH models.

Conclusion

Our study unveils a pharmacologically inducible switch wherein RSV, via direct VDAC1 binding, redirects its function toward TRIM23-mediated K27-linked ubiquitination and mitophagy. The TRIM23-VDAC1 allosteric axis represents a novel therapeutic paradigm for neuroprotective intervention in DR and beyond.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12964-026-03105-6.

Introduction

Diabetic retinopathy (DR) is a leading cause of acquired blindness in the global working-age population [1]. While historically characterized as a microvascular disorder, growing evidence positions retinal neurodegeneration as a critical early event in DR pathogenesis [2, 3]. The loss of retinal ganglion cells (RGCs) precedes and potentially exacerbates subsequent microvascular damage, highlighting the urgency of developing neuroprotective strategies. Given their high metabolic demands, RGCs are highly vulnerable to mitochondrial dysfunction under diabetic conditions [4, 5]. Disruption of mitochondrial homeostasis leads to the accumulation of damaged organelles, driving oxidative stress and initiating apoptotic pathways. Therefore, the efficient clearance of impaired mitochondria is essential for RGCs survival.

Voltage-dependent anion channel 1 (VDAC1), the most abundant protein on the outer mitochondrial membrane, functions as a crucial gatekeeper for mitochondrial permeability, thereby playing a central role in both apoptosis and mitochondrial quality control [6, 7]. Notably, ubiquitination of VDAC1 has been established as a central signal for recruiting autophagy adapters such as ubiquitin-binding protein p62 (p62) and microtubule-associated protein 1 light chain 3 (LC3), thereby directing damaged mitochondria to the lysosome for degradation [8, 9]. Its upregulation serves as a critical checkpoint in the pathogenesis of multiple neurodegenerative diseases, including Alzheimer’s disease, Parkinson’s disease, and amyotrophic lateral sclerosis [10, 11]. In the context of retinal trauma, elevated VDAC1 has been similarly shown to induce RGCs death through mechanisms involving calcium dysregulation and oxidative stress [12]. Consequently, inhibiting VDAC1 through small molecules, genetic knockdown, or peptide inhibitors, has emerged as a validated neuroprotective strategy [13–15].

The plant-derived polyphenol resveratrol (RSV) is a known protein ligand with multifaceted effects [16, 17]. Multiple studies have implicated VDAC1 as a potential therapeutic target of RSV, evidenced by its ability to inhibit VDAC1-mediated mitochondrial dysfunction in Parkinson’s disease models, promote VDAC1 deacetylation in cardiomyocytes, and reduce VDAC1 expression in neurodegenerative contexts [18–21]. These findings suggest a compelling, yet mechanistically unresolved, interaction between RSV and VDAC1.

The function of VDAC1 is tightly regulated by post-translational modifications, among which ubiquitination is particularly important for targeting damaged mitochondria for autophagic clearance [9, 22]. However, the specific E3 ubiquitin ligase that catalyzes VDAC1 ubiquitination in the DR remains unknown. Given that VDAC1 ubiquitination serves as a recognition signal for mitophagy receptors, identifying the responsible E3 ligase is critical for understanding how mitochondrial quality control is orchestrated in stressed RGCs. Tripartite motif-containing 23 (TRIM23) is an E3 ubiquitin ligase that acts as a positive regulator of selective autophagy [23]. Recent studies have further established TRIM23 as a critical node linking innate immunity to autophagy, functioning downstream of the cGAS-STING pathway to orchestrate autophagic responses to cellular stress [24]. These observations led us to investigate whether TRIM23 could represent a previously unexplored upstream regulator that confers specificity to VDAC1-mediated mitophagy, particularly in the context of RSV action.

In this study, we found that VDAC1 is related to the degeneration of RGCs in diabetic conditions, identified RSV as a potential VDAC1 inhibitor. We further demonstrate that RSV promotes the formation of a complex between TRIM23 and VDAC1, leading to the K27-linked ubiquitination of VDAC1. This specific molecular modification reprograms VDAC1 to a pro-mitophagy role, facilitating the clearance of damaged mitochondria and ultimately protecting RGCs from degeneration.

Materials and methods

Human donor sample collection

Posterior ocular tissues were obtained from 8 body donors (4 diabetic and 4 non-diabetic) through the Wuhan Red Cross Medical Center (Wuhan, China) under approved ethical guidelines (Ethical Approval: TJ-IRB20230899). All donors provided written informed consent prior to death, and samples were collected within 4–6 h postmortem to ensure tissue integrity.

Animal models and functional assessments

Male C57BL/6J mice (8 weeks old) were housed in the Specific Pathogen-Free (SPF) facility of the Experimental Animal Center, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology. The animals were maintained at 25 °C under a 12-h light/dark cycle with free access to standard diet and water. All procedures were approved by the Institutional Animal Care and Use Committee of the same institution (Ethical Approval No.TJH-202311038).

Streptozotocin (STZ)-induced diabetic model

Diabetes was induced by intraperitoneal injection of STZ. Briefly, STZ (Sigma-Aldrich, S0130) was dissolved in 0.1 M citrate-citrate sodium buffer (pH 4.5) to prepare a 1% solution. After overnight fasting, mice received an injection of STZ (70 mg/kg body weight) for two consecutive days. Fasting blood glucose ≥ 11.1 mmol/L at 72 h post-injection defined diabetic mice for inclusion. Diabetic mice received RSV (50 mg/kg/day) via oral gavage for 12 weeks, starting 3 days after STZ injection to allow stable hyperglycemia development.

Acute ocular hypertension (AOH) Model

AOH was induced by anterior chamber saline injection. Briefly, a 30-gauge needle connected to a pressure-controlled perfusion system was inserted into the anterior chamber under anesthesia. Sterile saline was infused to maintain intraocular pressure at 110 mmHg for 60 min, with continuous pressure monitoring. Successful ischemia was confirmed by pupillary dilation and loss of corneal reflex during maintenance, followed by reperfusion upon needle withdrawal. Mice were pretreated with RSV (100 mg/kg/day) for 2 days prior to AOH induction, followed by continued administration for 5 days.

Dose-response experiments of resveratrol

To determine optimal RSV concentrations, dose-response assays were performed in vitro and in vivo. SH-SY5Y cells were treated with 0, 2.5, 5, or 10 µM RSV under high glucose (HG) for 24 h, and apoptosis was assessed by TUNEL. Based on these in vitro results, 5 µM RSV was selected for subsequent cell experiments. STZ-induced diabetic mice received RSV at 10, 50, or 100 mg/kg/day by oral gavage for 12 weeks, and AOH mice received the same doses for 5 days. Retinal function was evaluated by ERG. Based on the results, 50 mg/kg/day was selected for the chronic STZ model and 100 mg/kg/day for the acute AOH model.

AAV preparation and intraocular injections

Adeno-associated virus serotype 2 (AAV2) carrying Trim23 under the control of the γ-synuclein (SNCG) promoter (AAV2-SNCG-Trim23) was generated as previously described. For intravitreal delivery, adult mice were anesthetized and injected with 1 µL of AAV2-SNCG-Trim23 (2 × 10¹² genome copies/mL) using a pulled glass micropipette under a dissecting microscope to ensure precise administration. Post-injection, 1% atropine eye drops were applied to prevent pupillary constriction, followed by topical application of ofloxacin gel to reduce infection. All procedures were performed under dim light to reduce retinal phototoxicity and ensure optimal viral transduction efficiency. AAV2 virus was purchased from AuGCT (Wuhan, China).

Immunofluorescence staining

Paraffin-embedded mouse retinal Sect.  (4 µm) were deparaffinized in xylene and rehydrated through graded ethanol series. Antigen retrieval was performed using citrate buffer (pH 6.0) under high-pressure conditions (7 min). After blocking with 5% normal goat serum for 1 h at room temperature, sections were incubated overnight at 4°C with primary antibodies against VDAC1 (Abcam, ab15895, 1:200) and RBPMS (Proteintech, 15187-1-AP, 1:200). Following PBST washes, sections were incubated with species-matched Alexa Fluor-conjugated secondary antibodies (Jackson immunoresearch, 705-585-003, 1:200) for 1 h at room temperature. Nuclei were counterstained with 4’,6-diamidino-2-phenylindole (DAPI) (Servicebio, G1012, 5 min), and slides were mounted with anti-fade medium (Servicebio, G1401). Fluorescence images were acquired using a confocal laser scanning microscope (Zeiss, LSM 900) with consistent exposure settings across samples.

Fluorescence in situ hybridization (FISH) for Vdac1 mRNA in retinal sections

FISH was performed using the RNAscope^® technology-based assay. Paraffin-embedded mouse eye sections were deparaffinized in xylene and rehydrated through a graded ethanol series. Antigen retrieval was performed by heating sections in citrate buffer (pH 6.0) at 100 °C for 15 min. Sections were then incubated with proteinase K (20 µg/mL) at 40 °C for 5 min to enhance tissue permeability. Custom-designed Z-pair probes targeting mouse Vdac1 mRNA were synthesized (p-m-vdac1 probe set; 500 nM). Hybridization was performed overnight at 40 °C in hybridization buffer containing the Vdac1 probe set. Following hybridization, sequential signal amplification was achieved by incubating sections with pre-amplifier (40 °C, 45 min) followed by amplifier and Cy3-conjugated label probes (40 °C, 3 h). Nuclei were counterstained with DAPI (Servicebio, G1012) for 8 min. Fluorescence images were captured using a Nikon Eclipse ci fluorescence microscope. All washing steps were performed in RNase-free conditions using DEPC-treated solutions.

Hematoxylin and eosin (HE) staining

Tissue sections were deparaffinized in xylene (2 × 5 min) and rehydrated through a graded ethanol series (100% → 70%). Nuclei were stained with Harris hematoxylin (2 min), followed by differentiation in 1% acid alcohol (5–10 s) and bluing in 0.084% ammonia water. Cytoplasmic counterstaining was performed using eosin (1 min), after which sections were dehydrated in ethanol, cleared in xylene, and mounted with neutral resin. For optimal results, staining intensity was microscopically monitored during differentiation to ensure clear nuclear contrast without over-decolorization.

Electroretinography (ERG)

Full-field ERGs were recorded using a RetiMINER system (IRC, Chongqing, China). Mice were dark-adapted for ≥ 12 h, then anesthetized with pentobarbital sodium via intraperitoneal injection, and pupils were dilated with Medolite eye drops. ERGs were elicited by flash stimuli ranging from 0.01 to 10 cd·s/m² under dark-adapted conditions. Oscillatory potentials (OPs) were isolated from the ERG waveform at 3.0 cd·s/m². Signals were acquired with corneal electrodes, and responses were averaged across multiple trials to minimize noise. All procedures adhered to institutional guidelines for animal electrophysiology and were consistent with established ERG recording standards.

Transmission electron microscopy (TEM)

Mice were transcardially perfused with 2.5% glutaraldehyde + 2% paraformaldehyde in 0.1 M phosphate buffer (PB, pH 7.4). After enucleation, eyecups were post-fixed in the same fixative for 24 h at 4 °C. Central retinal pieces (1 × 1 mm²) containing the ganglion cell layer were dissected, post-fixed in 1% OsO₄ with 1.5% potassium ferrocyanide (2 h, 4 °C), block-stained with 1% uranyl acetate (1 h), and embedded in Epon 812 after graded ethanol-acetone dehydration. Ultrathin Sect.  (70 nm) were cut using a Leica EM UC7 ultramicrotome, contrasted with uranyl acetate and lead citrate, and examined with a Hitachi HT7800 TEM at 80 kV. Mitochondrial morphology and mitophagosomes in RGCs were evaluated.

Cell models and functional assays

Cell lines and culture conditions

The SH-SY5Y cell line (Cell Health Biotechnology, Wuhan, China) and R28 retinal precursor cells (Aopei Biotechnology, Shanghai, China) were maintained in DMEM (Gibco, 11885084) supplemented with 10% fetal bovine serum (Cegrogen, A0500-3010), 100 U/mL penicillin G, and 100 mg/mL streptomycin sulfate (Servicebio, G4003) at 37 °C in a 5% CO₂ humidified incubator.

For HG treatment, cells were exposed to 25 mM glucose for 24 h. For oxygen-glucose deprivation/reoxygenation (OGD/R) treatment, cells were cultured in glucose-free DMEM under hypoxic conditions (1% O₂, 5% CO₂, 94% N₂) for 6 h, followed by reoxygenation in normal DMEM (5mM glucose, 10% FBS) under normoxia for 18 h. Resveratrol (RSV) was used at 5 µM.

Immunoprecipitation and western blotting

Cells were lysed in ice-cold RIPA buffer (Beyotime, P00138) supplemented with protease and phosphatase inhibitors. Lysates were centrifuged (12,000 ×g, 15 min, 4 °C), and supernatants were pre-cleared with Protein A/G magnetic beads (MCE, HY-K0202). For immunoprecipitation, cell lysates were simultaneously incubated with anti-Flag (Proteintech, 66008-4-Ig/20543-1-AP, 1:200), anti-Myc (Proteintech, 10828-1-AP, 1:1000), or anti-HA (CST, 3537, 1:100) antibodies and protein A/G magnetic beads in a rotating mixer overnight at 4 °C. Beads were washed four times with lysis buffer, and bound proteins were eluted in 2× Laemmli buffer (95 °C, 5 min) for subsequent Western blot analysis.

Proteins were resolved by SDS-PAGE and transferred to PVDF membranes. Membranes were blocked with 5% non-fat milk in TBST (1 h, RT) and probed overnight (4 °C) with primary antibodies against VDAC1 (Abcam, ab15895, 1:1000), ACTB (Abclonal, AC009, 1:5000), LC3B (CST, 83506 S, 1:1000), p62 (CST, 39749 S, 1:2000), TRIM23 (Abcam, ab192032, 1:1000), Flag (Proteintech, 20543-1-AP, 1:2000), Ubiquitin (CST, 3936 S, 1:1000), HA (CST, 3537, 1:2000), or Myc (Proteintech, 10828-1-AP, 1:2000). HRP-conjugated secondary antibodies (Abclonal, AS014/AS003/AS061, 1:10000) were applied, and signals were detected using enhanced chemiluminescence (Biosharp, BL520A).

Total RNA isolation and real-time PCR

Total RNA was isolated from SH-SY5Y cells using the Super FastPure Cell RNA Isolation Kit (Vazyme, RC102-01) according to the manufacturer’s protocol. RNA concentrations were measured using a NanoDrop 2000 spectrophotometer (Thermo Scientific). cDNA was synthesized using the HiScript III RT SuperMix for qPCR (+ gDNA wiper) (Vazyme, R323-01), which contains a gDNA removal module to eliminate residual genomic DNA contamination. Quantitative real-time PCR was subsequently carried out on a Real-Time PCR System (Roche, LC96) using the Taq Pro Universal SYBR qPCR Master Mix (Vazyme, Q712-02). The mRNA levels were quantified by normalizing to β-actin. The following primer sequences were used: VDAC1 forward, 5’-ACGTATGCCGATCTTGGCAAA-3’; VDAC1 reverse, 5’-TCAGGCCGTACTCAGTCCATC-3’; ACTB forward, 5’-CATGTACGTTGCTATCCAGGC-3’; ACTB reverse, 5’-CTCCTTAATGTCACGCACGAT-3’.

Plasmid transfection

The HEK-293T embryonic kidney cell line (Shanghai Zhong Qiao Xin Zhou Biotechnology) was maintained in high-glucose DMEM (Gibco, C11995500BT) supplemented with 10% fetal bovine serum (Cegrogen, A0500-3010), 100 U/mL penicillin G, and 100 mg/mL streptomycin sulfate (Servicebio, G4003) at 37 °C under 5% CO₂. For transfection experiments, cells at 70–80% confluence were transfected with plasmid DNA using HighGene plus Transfection Reagent (Abclonal, RM09014P) following the manufacturer’s protocol. Post-transfection, cells were cultured for an additional 48 h before downstream analysis to ensure sufficient transgene expression.

siRNA transfection

SH-SY5Y cells were seeded in 24-well plates at a density of 1 × 10⁵ cells per well. EGFP-tagged siRNA (20 pmol per well) was diluted in 200 µL Opti-MEM reduced-serum medium, mixed gently with Lipofectamine™ RNAiMAX transfection reagent (1.5 µL per well), and incubated at room temperature for 20 min to allow complex formation. The following siRNA duplexes were used: non-targeting control, sense 5´-UUCUCCGAACGUGUCACGUTT-3´ and antisense 5´- ACGUGACACGUUCGGAGAATT − 3´; human VDAC1-specific siRNA, sense 5´-GGAUACACUCAGACUCUAAAGTT-3´ and antisense 5´- CUUUAGAGUCUGAGUGUAUCCTT-3´; human TRIM23-specific siRNA, sense 5´- AGCUGUUGUGUUUGUUGUAGATT-3´ and antisense 5´- UCUACAACAAACACAACAGCUTT-3´. At 72 h post-transfection, EGFP fluorescence was visualized using a fluorescence microscope, and cells were lysed in RIPA buffer supplemented with protease inhibitors, and protein concentration was determined using a BCA assay.

Measurement of mitochondrial membrane potential

Mitochondrial membrane potential (ΔΨm) was measured in SH-SY5Y cells using JC-10 fluorescent probe (Yeasen, 40752ES60). Cells were seeded in glass-bottom dishes (1 × 10⁵ cells/well), treated for 24 h, then incubated with 10 µM JC-10 at 37 °C for 20 min. In healthy mitochondria, JC-10 forms red fluorescent aggregates (Ex = 540/Em = 590 nm), while depolarized mitochondria show green fluorescence (Ex = 490/Em = 525 nm) due to monomeric dispersion. Fluorescence was captured by fluorescence microscopy (Olympus BX53 Series Microscope), and ΔΨm was quantified as the red/green fluorescence ratio using ImageJ.

Biophysical validation of resveratrol as a VDAC1 ligand

Structure-based virtual screening

A structure-based virtual screening was performed to identify potential ligands targeting human VDAC1 (PDB ID: 6G6U). The crystal structure was prepared using the Protein Preparation Wizard module in Schrödinger Maestro, which included the addition of hydrogens and energy minimization. The primary binding site was defined based on the top-ranked site predicted by the software, and a receptor grid was generated with dimensions of 20 Å × 20 Å × 20 Å.

The HY-L021 Natural Product Library (MedChemExpress, Monmouth Junction, NJ, USA), comprising approximately 4,820 compounds, was used for virtual screening. Three-dimensional structures of all compounds were obtained from the supplier and prepared for docking using the LigPrep module in Schrödinger Maestro. The screening was conducted using the Glide module through a sequential workflow: an initial high-throughput virtual screening (HTVS) was followed by standard precision (SP) and extra precision (XP) docking rounds, retaining the top 15% of compounds from each preceding round for the next. The final ranked list of compounds was visually inspected, and resveratrol was selected as a top candidate for further experimental validation.

Expression and purification of recombinant VDAC1 protein

The recombinant VDAC1 protein was expressed in Escherichia coli using the pColdⅡ vector system. The codon-optimized VDAC1 gene was transformed into competent E. coli cells. After cell lysis by sonication, the inclusion bodies were washed and solubilized in denaturing buffer (8 M urea, 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 20 mM imidazole). The protein was purified under denaturing conditions using Ni-NTA affinity chromatography, followed by refolding via dialysis into a native buffer system. Further purification was performed by size-exclusion chromatography (Superdex 200), The final protein preparation was confirmed to be > 99% pure by SDS-PAGE, and its identity was verified by Western blot (using an anti-VDAC1 antibody) and LC-MS analysis, the latter yielding a molecular weight of 31,770.1016 Da.

Surface Plasmon Resonance (SPR) spectroscopy

The SPR experiments were performed using an OpenSPR™ instrument (Nicoya) equipped with a COOH sensor chip to investigate the interaction between the ligand protein VDAC1 and the analyte RSV. The chip was activated with EDC/NHS (1:1) solution, followed by immobilization of 200 µL of VDAC1 diluted in 10 mM sodium acetate buffer (pH 5.0) at a flow rate of 20 µL/min. After blocking, RSV was diluted in 1% DMSO buffer and injected at varying concentrations with an association time of 240 s and a dissociation time of 360 s. The binding kinetics were monitored in real-time at 25 °C, and the data were analyzed using the TraceDrawer software (Ridgeview Instruments AB, Sweden) with a One-To-One binding model. The affinity constant (KD), association rate (ka), and dissociation rate (kd) were determined. All steps were conducted under standardized conditions to ensure reproducibility.

Mass spectrometry-protein identifications

The LC-MS/MS analysis was performed to identify VDAC1-interacting proteins under RSV treatment conditions. HEK-293T cells were transfected with flag-VDAC1 plasmid, then divided into two groups treated with or without RSV for 24 h. The immunoprecipitants were separated by SDS-PAGE and visualized by Coomassie Brilliant Blue staining. For mass spectrometry, the proteins were separated on SDS-PAGE gels. After in‐gel digestion, the proteins were analysed using a timsTOF Pro (Bruker Daltonics) coupled to UltiMate 3000 RSLCnano system (Thermo). The mass spectrometry data were analyzed using MaxQuant software (v2.0.3.0) with database searching against UniProt human database (release 2023_01) at a 1% false discovery rate, where differentially expressed proteins were identified based on spectral counts and statistical significance (p < 0.05).

Statistical analysis

Data were analyzed using GraphPad Prism 7.0 and presented as mean ± SEM (standard error of the mean). For comparisons between two groups, unpaired two-tailed Student’s t-tests were performed, while one-way ANOVA followed by Tukey’s post-hoc test was applied for multi-group comparisons. Statistical significance was defined as p < 0.05.

Results

VDAC1 is upregulated in diabetic retinopathy and its knockdown attenuates high glucose-induced apoptosis

Given the critical role of mitochondrial dysfunction in neurodegeneration, we sought to investigate the potential involvement of the mitochondrial gatekeeper protein VDAC1 in the context of DR. We began by assessing its expression profile across clinical samples, animal models, and cellular models. Immunofluorescence analysis of posterior ocular segments (with corneas removed) from 8 body donors (4 non-diabetic and 4 diabetic) obtained through the Wuhan Red Cross Medical Center. The donor details, including average ages of 66.5 years for the non-diabetic (NR) group and 65.25 years for the diabetic (DR) group, are provided in Table 1. Immunofluorescence staining of retinal sections from these donors revealed a significant upregulation of VDAC1 expression in the RGCs of the DR group compared to the NR group (Fig. 1A, B).

Table 1.

Donor information for human retinal tissues

Donor ID	Sex	Age (years)	Diabetic Status	Cause of Death
1	Male	82	No	Myocardial Infarction; Pulmonary Infection
2	Male	67	No	Gastric Cancer; Respiratory Failure
3	Female	73	No	Respiratory Failure
4	Female	44	No	Breast Cancer with Bone Metastasis
5	Male	71	Yes	Respiratory Failure
6	Male	76	Yes	Myocardial Infarction
7	Female	70	Yes	Pancreatic Cancer
8	Male	44	Yes	Diabetic Ketoacidosis

Fig. 1.

VDAC1 is Upregulated in RGCs in Diabetes and Mediates High Glucose-Induced Apoptosis. A, B VDAC1 immunofluorescence (A) and its quantification (B) in the GCL of human donor retinas. Scale bar: 20 μm. C, D Representative images (C) and quantification (D) of VDAC1 immunofluorescence in the retinas of STZ-induced diabetic mice. Scale bar: 20 μm. E, F Western blot analysis (E) and quantitative summary (F) of VDAC1 protein levels in mouse retinas and HG-treated SH-SY5Y cells. G, H Efficiency of VDAC1 knockdown in SH-SY5Y cells, confirmed by Western blot (G) and its quantification (H). I, J TUNEL staining (I) and quantitative analysis (J) of apoptotic cells under indicated conditions. Scale bar: 50 μm. Data are mean ± SEM (n = 3–5). *P < 0.05, **P < 0.01, ***P < 0.001

To determine if this observation is recapitulated in experimental models, we examined the retinas of STZ-induced diabetic mice. Consistent with the human data, a pronounced increase in VDAC1 immunofluorescence was observed in the GCL of diabetic mice (Fig. 1C, D). Furthermore, this upregulation was consistently observed at the protein level in both retinal tissues from diabetic mice and in HG-treated SH-SY5Y cells, a well-established in vitro model for neuronal studies, as determined by Western blot analysis (Fig. 1E, F).

Furthermore, to determine the functional significance of VDAC1 upregulation, we performed knockdown experiments using siRNA targeting VDAC1 in SH-SY5Y cells. Efficient knockdown of VDAC1 (Fig. 1G, H, approximately 54.9% reduction) significantly attenuated HG-induced apoptosis (Fig. 1I, J), indicating that VDAC1 is not merely a biomarker but a key mediator of RGCs injury under diabetic conditions. Thus, VDAC1 upregulation is a pathogenic event in diabetic retinas, and its suppression is sufficient to alleviate HG-induced neuronal apoptosis.

Identification and validation of resveratrol as a VDAC1-binding compound that promotes its degradation

Having established that VDAC1 knockdown is sufficient to mitigate HG-induced apoptosis, we next sought to identify a small-molecule agent capable of downregulating VDAC1 protein levels for therapeutic exploration. To this end, we performed a structure-based virtual screening of a natural product library, with a specific focus on compounds known to target mitochondrial pathways. The structural basis (VDAC1 3D structure) and procedural workflow of the virtual screening are detailed in Supplementary Fig. 1A, B, respectively. Among the top-ranked molecules from our virtual screen for mitochondrial therapeutics, RSV exhibited high predicted binding affinity for VDAC1, forming interactions with residues N76, D100, and K109 (Fig. 2A). Our screening result is consistent with prior studies reporting RSV’s ability to modulate mitochondrial VDAC1, thereby strengthening its candidacy for further experimental validation [19].

Fig. 2.

Resveratrol binds to VDAC1 and attenuates its diabetic upregulation via a post-transcriptional mechanism. A Structure-based virtual screening identified potential VDAC1 interactors. Views include the protein surface, the detailed binding pose of the ligand, and key receptor-ligand interactions. B Concentration-dependent binding curve of VDAC1 with RSV measured by SPR. C Direct interaction between RSV and VDAC1 demonstrated by pull-down assay using RSV-conjugated beads. D Western blot analysis of VDAC1 protein levels in retinas from STZ-induced diabetic mice (CON, STZ, STZ + RSV) and in HG-stimulated SH-SY5Y cells (CON, HG, HG + RSV). E Quantification of the data shown in (D). F, G Representative immunofluorescence images (F) and quantitative analysis (G) of VDAC1 (green) in the RGCs of STZ-induced diabetic mice. Nuclei are stained with DAPI (blue). Scale bar: 20 μm. H, I Detection of Vdac1 mRNA (H) and its quantification (I) in retinal sections from CON, STZ, and STZ + RSV treated mice. Scale bar: 20 μm. J Quantitative RT-PCR analysis showing VDAC1 mRNA levels in SH-SY5Y cells under CON, HG and HG + RSV treatment. Data are mean ± SEM (n = 3–5). *P < 0.05, **P < 0.01, ***P < 0.001

To experimentally validate this computational prediction, we purified recombinant VDAC1 protein and conducted SPR analysis. To obtain the recombinant VDAC1 protein for binding assays, the protein was expressed in E. coli, purified under denaturing conditions using Ni-NTA affinity chromatography, and subsequently refolded by dialysis. The final preparation had a concentration of 0.8039 mg/mL and a purity exceeding 99%, as confirmed by SDS-PAGE (Supplementary Fig. 2A) and Western blot analysis (Supplementary Fig. 2B). The identity and molecular weight of the protein were unequivocally verified by mass spectrometry, which yielded a measured molecular mass of 31,770.1016 Da, consistent with the theoretical value. The corresponding mass spectrometric profiles, including the total ion chromatogram, UV trace, source spectrum, and deconvoluted spectrum, are provided in Supplementary Fig. 2C-E. SPR analysis confirmed a direct interaction between RSV and VDAC1, with steady-state fitting yielding a dissociation constant of 3.02e-5 M, indicative of moderate affinity, and a clear concentration-dependent binding response (Fig. 2B and Supplementary Fig. 2F). To confirm this binding in a more biologically relevant context, we performed a pull-down assay using RSV-conjugated magnetic beads. Western blot analysis of the pull-down product successfully detected VDAC1 from SH-SY5Y cell lysates, unequivocally demonstrating their direct interaction in a cellular environment (Fig. 2C).

To determine the optimal concentration of RSV for subsequent experiments, we performed dose-response assays in vitro and in vivo. In SH-SY5Y cells, RSV at 5 µM and 10 µM both significantly suppressed HG-induced apoptosis compared to HG alone, with no statistical difference between the two concentrations; 2.5 µM RSV was suboptimal (Supplementary Fig. 3A, B). In STZ-induced diabetic mice, oral RSV at 10, 50, or 100 mg/kg/day for 12 weeks revealed that 50 mg/kg/day provided the strongest protection against diabetes-induced declines in a-wave, b-wave, and OPs, while 10 mg/kg/day showed no significant improvement and 100 mg/kg/day exhibited mild toxicity (Supplementary Fig. 3C, D). Based on these results, 5 µM RSV was selected for in vitro mechanistic studies, and 50 mg/kg/day for the chronic STZ model.

We next evaluated the functional effect of RSV on VDAC1 in disease models. In both diabetic mouse retinas and HG-stimulated SH-SY5Y cells, RSV treatment significantly reduced the elevated VDAC1 protein levels (Fig. 2D, E). Consistently, immunofluorescence staining of retinal sections confirmed that RSV effectively reversed the STZ-induced upregulation of VDAC1 in the GCL of diabetic mice (Fig. 2F, G).

FISH on retinal sections revealed that Vdac1 mRNA in the GCL was significantly upregulated in STZ mice compared to controls; however, RSV treatment did not reverse this increase, as Vdac1 mRNA levels remained elevated in the STZ + RSV group (Fig. 2H, I). These findings collectively indicate that RSV does not suppress VDAC1 expression at the transcriptional level. Therefore, RSV regulates VDAC1 through a post-transcriptional mechanism, likely involving enhanced protein degradation rather than decreased mRNA transcription. Consistent with the in vivo observation that RSV does not suppress Vdac1 transcription, we next evaluated Vdac1 mRNA expression in vitro. In SH-SY5Y cells, quantitative RT-PCR analysis showed that RSV treatment did not significantly alter VDAC1 mRNA expression under HG conditions (Fig. 2J).

Resveratrol treatment preserves retinal structure and function in diabetic models

To exclude the possibility that RSV exerts non-specific effects on normal retina or cells, we treated healthy mice with RSV (50 mg/kg/day for 12 weeks) and SH-SY5Y cells with 5 µM RSV for 24 h. As shown in Supplementary Fig. 4, RSV alone did not alter retinal thickness (Supplementary Fig. 4A, B), RNA-binding protein with multiple splicing (RBPMS)⁺ RGC counts (Supplementary Fig. 4C, D), VDAC1 expression (Supplementary Fig. 4E, F, K, L), ERG function (Supplementary Fig. 4G, H), or apoptosis (Supplementary Fig. 4I, J) compared to vehicle controls. These data confirm that RSV alone does not exert non-specific effects on normal retina or SH-SY5Y cells under the tested conditions.Given the RSV-mediated downregulation of VDAC1, we next evaluated its functional consequences on retinal integrity and function in vivo. In diabetic mice, RSV treatment effectively preserved retinal architecture, as evidenced by HE staining showing a more organized retinal structure and quantitative analysis confirming the preservation of total retinal thickness (Fig. 3A, B). Functionally, scotopic ERG recordings demonstrated that RSV treatment significantly restored the amplitudes of OPs, a-waves, and b-waves, indicating improved retinal neural signaling (Fig. 3C, D). At the cellular level, immunofluorescence staining for the RGCs marker RBPMS and its subsequent quantification revealed that RSV significantly prevented the diabetes-induced loss of RGCs (Fig. 3E, F). Moreover, TUNEL staining of SH-SY5Y cells showed a marked reduction in apoptotic cells upon RSV treatment under HG conditions (Fig. 3G, H), confirming its anti-apoptotic efficacy in vitro. Furthermore, phase-contrast microscopy revealed that compared to control cells, HG stimulation significantly reduced neurite length in SH-SY5Y cells, and RSV treatment reversed this effect (Fig. 3I). Collectively, these data demonstrate that RSV confers comprehensive protection against diabetes-induced retinal neurodegeneration, preserving tissue architecture, neural function, and cellular viability.

Fig. 3.

Resveratrol ameliorates diabetes-induced RGCs injury in vivo and in vitro. A, B Retinal morphology assessed by HE staining (A), with quantitative analysis of total retinal thickness shown in (B). Scale bar: 200 μm. C, D ERG traces recorded under scotopic conditions (C) and quantitative analysis of the amplitudes of OPs, a-waves, and b-waves from the ERG recordings (D). E, F Immunofluorescence labeling of RGCs with RBPMS in retinal sections (E) and its corresponding quantification (F). Nuclei are stained with DAPI. Scale bar: 50 μm. G, H TUNEL staining (G) and quantitative analysis (H) of apoptotic cells. Scale bar: 50 μm. I Phase-contrast micrographs showing the morphological changes of SH-SY5Y cells. Scale bar: 50 μm. Data are mean ± SEM (n = 3–5). *P < 0.05, **P < 0.01, ***P < 0.001

Resveratrol activates mitophagy to clear damaged mitochondria

Given the central role of VDAC1 in mitochondrial homeostasis, along with our findings that RSV downregulates VDAC1 and ameliorates RGC injury under diabetic conditions, we hypothesized that RSV confers protection by mitigating mitochondrial damage. TEM of the RGCs (Fig. 4A) revealed that in diabetic mice, mitochondria exhibited prominent swelling and cristae disruption. RSV treatment not only ameliorated these pathological changes but also induced classic mitophagy features, including double-membrane autophagosomes sequestering damaged mitochondria. These observations provide direct ultrastructural evidence that RSV promotes mitophagy in diabetic RGCs.

Fig. 4.

Resveratrol promotes mitophagy in DR Models. A TEM images of the GCL from CON, STZ, and STZ + RSV mice. Low-magnification view (left) outlines the RGCs region, with high-magnification insets (right) revealing ultrastructural details of mitochondria. Scale bars: 10 μm (low mag), 200 nm (high mag). B, C Western blot analysis (B) and quantitative densitometry (C) of the autophagy flux markers LC3 and p62 in SH-SY5Y cells under HG conditions with or without RSV treatment. D Confocal microscopy images showing the co-localization of mitochondria (labeled with MitoTracker Green) and lysosomes (labeled with LysoTracker Red) in SH-SY5Y cells under CON, HG, and HG + RSV conditions. Scale bar: 20 μm. E, F Assessment of mitochondrial membrane potential in SH-SY5Y cells using JC-10 staining (E). Red fluorescence indicates healthy, polarized mitochondria, while green fluorescence indicates depolarized mitochondria. The ratio of red to green fluorescence intensity is quantified in (F). Scale bar: 50 μm Data are presented as mean ± SEM (n = 3–5). *P < 0.05, **P < 0.01, ***P < 0.001

To further characterize the molecular events underlying RSV-induced mitophagy, we examined autophagic flux in HG-stimulated SH-SY5Y cells. Compared to the CON group, HG stimulation significantly increased p62 expression and decreased the LC3-II/I ratio, indicating impaired autophagic flux under HG conditions. Importantly, RSV treatment effectively reversed these changes, as evidenced by reduced p62 levels and an increased LC3-II/I ratio in the HG + RSV group compared to HG alone (Fig. 4B, C). The initiation of this degradative pathway was further demonstrated by a significant increase in the co-localization of mitochondria with lysosomes upon RSV treatment, as shown by confocal microscopy (Fig. 4D). Crucially, the clearance of damaged mitochondria via mitophagy led to a functional improvement in the mitochondrial network, as evidenced by the restoration of mitochondrial membrane potential under HG conditions, measured by an increased red/green fluorescence ratio in JC-10 staining (Fig. 4E, F). These results demonstrate that RSV activates a complete mitophagic response to clear damaged mitochondria, which in turn underlies the structural and functional protection of the diabetic retina.

To further validate these findings in a retinal-specific context, we repeated key experiments in R28 cells, a rat retinal neuronal cell line. As shown in Supplementary Fig. 5, RSV treatment significantly reduced VDAC1 protein levels (Supplementary Fig. 5A, B) and suppressed apoptosis (Supplementary Fig. 5C, D) under HG conditions, while also promoting mitolysosome co-localization (Supplementary Fig. 5E). Moreover, in an OGD/R model that mimics AOH injury, RSV similarly downregulated VDAC1 expression (Supplementary Fig. 5F, G), reduced apoptosis (Supplementary Fig. 5H, I), and enhanced mitolysosome formation (Supplementary Fig. 5J) in R28 cells. These data confirm that the protective effects of RSV on VDAC1 regulation, apoptosis suppression, and mitophagy activation are recapitulated in retinal-specific cells under both metabolic and ischemic stress conditions.

Resveratrol promotes TRIM23-VDAC1 interaction and ubiquitin-dependent VDAC1 degradation

Our findings established that RSV binds to and downregulates VDAC1, thereby promoting mitophagy. However, the precise molecular mechanism by which RSV orchestrates this process remained unknown. To systematically identify the specific machinery, we conducted co-immunoprecipitation coupled with mass spectrometry (Co-IP/MS) in HEK-293T cells overexpressing flag-VDAC1 and treated with or without RSV. The efficiency of VDAC1 overexpression and the successful immunoprecipitation of the target protein were verified by Coomassie Brilliant Blue staining and Western blot analysis, respectively (Supplementary Fig. 6A, B). Gene Ontology (GO) analysis of the VDAC1 interactome revealed a significant enrichment of proteins implicated in post-translational modification upon RSV treatment (Fig. 5A).

Fig. 5.

TRIM23 interacts with and regulates the ubiquitination and stability of VDAC1. A GO analysis of VDAC1-interacting proteins immunopurified from cells treated with RSV. B Immunoblot analysis of polyubiquitinated VDAC1 following immunoprecipitation from cells treated with or without RSV. C, D Co-IP assay verifying the physical interaction between TRIM23 and VDAC1. E Ubiquitination immunoblots of VDAC1 in SH-SY5Y cells overexpressing TRIM23. F Western blot showing VDAC1 protein levels in SH-SY5Y cells overexpressing TRIM23 or vector control. G Co-IP and immunoblot analysis of the effect of RSV stimulation on the interaction between VDAC1 and TRIM23

Guided by this finding, we directly assessed the impact of RSV on VDAC1 modifications. Immunoblot analysis demonstrated that RSV treatment significantly enhanced the ubiquitination of VDAC1 (Fig. 5B), while having a minimal effect on its phosphorylation status (Supplementary Fig. 6C), pinpointing ubiquitination as the dominant regulatory mechanism.

To identify the specific E3 ubiquitin ligase responsible for VDAC1 ubiquitination, we re-analyzed our Co-IP/MS dataset focusing on E3 ligases. Among the candidates, members of the TRIM family, including TRIM21 and TRIM23, were detected. Given that TRIM proteins are established regulators of selective autophagy and mitophagy, we focused on these candidates for further validation [25]. We therefore performed independent co-immunoprecipitation assays to validate whether these candidates indeed bind to VDAC1. Subsequent co-immunoprecipitation experiments revealed that TRIM23, but not TRIM21, specifically binds to VDAC1 (Fig. 5C, D and Supplementary Fig. 6D).

Having identified TRIM23 as a specific VDAC1-interacting E3 ligase, we then investigated its functional impact on VDAC1. Overexpression of Myc-TRIM23 significantly enhanced the ubiquitination of VDAC1 (Fig. 5E). Consistent with this ubiquitin ligase activity, graded overexpression of Myc-TRIM23 in SH-SY5Y cells treated with cycloheximide (CHX) led to a corresponding decrease in VDAC1 protein levels, demonstrating that TRIM23 promotes VDAC1 degradation (Fig. 5F).

The interaction between VDAC1 and TRIM23 was further visualized in SH-SY5Y cells, where immunofluorescence staining demonstrated clear co-localization of transfected flag-VDAC1 and Myc-TRIM23 (Supplementary Fig. 6E). Furthermore, we confirmed that RSV treatment robustly enhances this endogenous TRIM23-VDAC1 interaction in SH-SY5Y cells (Fig. 5G), without altering the expression level of TRIM23 itself (Supplementary Fig. 6F, G).

We next examined whether diabetes affects endogenous TRIM23 expression in the retina. Immunofluorescence staining of retinal sections from control and STZ-induced diabetic mice revealed that TRIM23 is broadly expressed across all retinal layers. In control mice, strong TRIM23 signals were readily detected in RGCs. Notably, in STZ mice, TRIM23 fluorescence intensity in the GCL was reduced compared to controls (Supplementary Fig. 6H, I). This indicates that diabetic conditions downregulate TRIM23 in RGCs.

TRIM23 catalyzes K27-linked ubiquitination of VDAC1 via specific domain interaction

To map the precise binding interfaces critical for the TRIM23-VDAC1 complex formation, we constructed a series of deletion mutants based on their functional domains. TRIM23 contains a canonical N-terminal RING domain, which confers E3 ubiquitin ligase activity, and a C-terminal ARF domain, which is involved in protein-protein interactions and membrane association. VDAC1, on the other hand, possesses an N-terminal α-helix that is known to regulate its channel gating and interaction with partner proteins. We hypothesized that this flexible region might serve as a docking site for TRIM23. To dissect their respective roles, we generated plasmids encoding full-length TRIM23 (WT), a RING domain deletion mutant (ΔRING), and an ARF domain deletion mutant (ΔARF), alongside VDAC1 WT and an N-terminal deletion mutant (VDAC1 Δ1–25) (Fig. 6A). Co-immunoprecipitation assays revealed that while the ΔARF mutant retained binding, the ΔRING mutant completely lost its ability to interact with VDAC1 (Fig. 6B). Similarly, the VDAC1 Δ1–25 mutant was severely compromised in its interaction with TRIM23 (Fig. 6C). These results unequivocally demonstrate that the RING domain of TRIM23 and the N-terminal α-helix of VDAC1 are indispensable for their mutual binding.

Fig. 6.

Mapping the TRIM23-VDAC1 interaction interface and defining the K27-linked ubiquitination mechanism. A Schematic diagrams depicting the amino acid ranges of the WT and various mutant peptides of TRIM23 and VDAC1 used in this study. B Co-IP and immunoblotting analysis of the interaction between VDAC1 and TRIM23 WT or its deletion mutants (ΔARF, ΔRING). C Co-IP and immunoblotting analysis of the interaction between TRIM23 and VDAC1 WT or its N-terminal deletion mutant (Δ1–25). D Schematic flowchart of the experiment designed to verify the type of ubiquitin linkage on VDAC1 induced by TRIM23. E Determination of the linkage type of TRIM23-mediated VDAC1 ubiquitination by Co-IP and western blot

Having established the physical interaction, we sought to determine the functional outcome. We designed a strategy to identify the specific ubiquitin linkage type catalyzed by TRIM23 on VDAC1 (Fig. 6D). HEK-293T cells were co-transfected with plasmids encoding flag-VDAC1, Myc-TRIM23, and wild-type (WT) or the indicated lysine-only mutant ubiquitin (Ub). VDAC1 was immunoprecipitated with an anti-flag antibody and analyzed by immunoblotting with an anti-Ub antibody. We found that TRIM23 specifically mediates K27-linked polyubiquitination of VDAC1, as evidenced by the strong ubiquitin signal only in the presence of K27-only ubiquitin (Fig. 6E). These data delineate a precise molecular mechanism wherein RSV facilitates the recruitment of TRIM23 to the N-terminus of VDAC1, leading to its K27-linked polyubiquitination.

The TRIM23-VDAC1 axis is indispensable for RSV-induced mitophagy and cytoprotection

To directly link this molecular mechanism to the observed functional phenotypes, we assessed the necessity of the TRIM23-VDAC1 axis in RSV’s effects. We silenced TRIM23 expression in SH-SY5Y cells using specific siRNA, achieving efficient knockdown (Fig. 7A, B). Crucially, TRIM23 deficiency abolished the ability of RSV to reduce VDAC1 protein levels and activate autophagy, as shown by the blunted changes in LC3-II and p62 (Fig. 7C, D). Consistent with this finding, the RSV-induced enhancement of VDAC1 ubiquitination was also abrogated upon TRIM23 knockdown, directly linking TRIM23 to the key post-translational modification (Fig. 7E). This disruption at the molecular level translated to a failure in executing mitophagy. The robust increase in mitolysosomal co-localization, a hallmark of complete mitophagy, observed with RSV treatment alone was entirely lost in TRIM23-silenced cells (Fig. 7F). The functional consequence of this mechanistic uncoupling was a reversal of RSV’s anti-apoptotic effect. The number of TUNEL-positive apoptotic cells in the HG + RSV+si TRIM23 group returned to the high levels observed in the HG group (Fig. 7G, H). These results unequivocally establish that TRIM23 is not merely associated but is an essential mediator that couples RSV binding to the activation of the VDAC1 ubiquitination-mitophagy pathway, ultimately conferring cytoprotection.

Fig. 7.

TRIM23 knockdown abrogates resveratrol-induced mitophagy and cytoprotection in SH-SY5Y cells. A, B Western blot analysis (A) and quantification (B) confirming efficient knockdown of TRIM23 protein by siRNA. C, D Immunoblot analysis of VDAC1 protein levels and autophagy flux markers (LC3, p62) under indicated conditions (C), with quantitative summaries shown in (D). E Co-IP and immunoblot analysis of VDAC1 ubiquitination levels treated with TRIM23 knockdown. Scale bar: 20 μm. F Confocal microscopy images showing the co-localization of mitochondria (MitoTracker Green) and lysosomes (LysoTracker Red) in SH-SY5Y under the indicated conditions. Scale bar: 20 μm. G, H TUNEL staining (G) and quantification of apoptotic cells (H) in SH-SY5Y cells under the indicated treatments. Scale bar: 50 μm. Data are mean ± SEM (n = 3–5). *P < 0.05, **P < 0.01, ***P < 0.001

Finally, we evaluated the translational potential of enhancing the TRIM23-VDAC1 pathway in vivo. We hypothesized that combining RSV with RGCs-specific overexpression of Trim23 would yield superior neuroprotection by maximally activating this endogenous protective pathway. In STZ-induced diabetic mice, we implemented a combined treatment regimen (Fig. 8A): oral RSV administration was initiated 3 days after diabetes confirmation (blood glucose ≥ 11.1 mmol/L) and continued throughout the study. To achieve RGCs specific overexpression, we intravitreally injected an AAV2 vector encoding Trim23 under the control of the SNCG promoter, which we selected for its high specificity and efficacy in targeting RGCs. This AAV2 injection was performed 60 days post-STZ induction, allowing for a 30-day period of Trim23 overexpression prior to final assessment. This combination therapy did not further preserve retinal thickness compared to RSV alone, as quantified by total retinal thickness measurements (Fig. 8B, C). Functionally, scotopic ERG showed that the combination therapy most effectively restored the amplitudes of OPs, a-waves, and b-waves (Fig. 8D, E). Most importantly, immunohistochemical analysis and quantification of RBPMS-positive RGCs confirmed that the combined intervention provided the strongest protection against RGCs loss (Fig. 8F, G).

Fig. 8.

Combined treatment with resveratrol and Trim23 overexpression synergistically protects against RGCs neurodegeneration in DR and AOH models. A Experimental timeline for STZ model treated with RSV and AAV-SNCG-mediated Trim23 overexpression. B, C Representative HE-stained retinal sections (B) and quantitative analysis of total retinal thickness (C) from the indicated groups. Scale bar: 200 μm. D, E Scotopic ERG traces (D) and quantitative analysis of OPs, a-wave, and b-wave amplitudes (E). F, G Immunofluorescence labeling of RGCs with RBPMS in retinal sections (F) and corresponding quantification of RBPMS+ cells (G). Nuclei are stained with DAPI. Scale bar: 50 μm. H Experimental timeline for the AOH model treated with RSV and AAV-SNCG-mediated Trim23 overexpression. (I) Representative images of RGCs labeled by RBPMS in retinal whole mounts from AOH model mice. Scale bar: 100 μm. J, K Scotopic ERG traces (J) and quantitative analysis of OPs, a-wave, and b-wave amplitudes (K) from AOH model mice. Data are mean ± SEM (n = 3–5). *P < 0.05, **P < 0.01, ***P < 0.001

To assess the broader applicability of this strategy beyond chronic metabolic stress, we employed an AOH model, a well-established and rapid paradigm for inducing RGCs injury that is independent of diabetic pathology. To determine the optimal RSV dose for the acute AOH model, we first performed dose-response assays in AOH mice treated with 10, 50, or 100 mg/kg/day RSV. ERG analysis revealed that 50 mg/kg/day and 100 mg/kg/day both significantly improved a-wave, b-wave, and OP amplitudes compared to AOH alone, with 100 mg/kg/day showing a more robust protective effect, while 10 mg/kg/day showed no significant improvement (Supplementary Fig. 7A, B). Based on these results, 100 mg/kg/day was selected for subsequent AOH combination experiments.In this model (Fig. 8H), mice first received an intravitreal injection of AAV2-control or AAV2-SNCG-Trim23-EGFP to allow for 28 days of Trim23 expression in RGCs, followed by a 2-day oral RSV pretreatment. After AOH induction, RSV treatment was continued for another 3 days prior to the final assessment. The successful induction of retinal injury in the AOH model was confirmed by HE staining, which revealed a significant thinning of retinal thickness in the AOH group compared to the control group (Supplementary Fig. 7C, D). Strikingly, the combination of RSV and AAV2-SNCG-Trim23 also significantly protected RGCs in retinal whole mounts (Fig. 8I) and preserved retinal function in ERG recordings (Fig. 8J, K) in this model. Robust and RGC-specific overexpression of Trim23 was confirmed by co-immunofluorescence staining of EGFP (from the AAV2-SNCG-Trim23-EGFP vector) with the RGC marker RBPMS in retinal whole mounts (Supplementary Fig. 7E). These findings underscore the broad neuroprotective potential of therapeutically co-targeting VDAC1 and TRIM23, highlighting a promising combinatorial strategy for treating RGCs degeneration across distinct disease etiologies.

Discussion

The progression of DR is intrinsically linked to the early degeneration of RGCs, a process driven by mitochondrial dysfunction [2, 3]. Here, we identify the mitochondrial gatekeeper protein VDAC1 as a pivotal mediator of this pathology and delineate a complete pharmacologic pathway through which RSV counteracts it. We demonstrate that VDAC1 is upregulated in diabetic RGCs and is essential for hyperglycemia-induced apoptosis. Crucially, we define a novel mechanism wherein RSV directly binds VDAC1, recruits the E3 ubiquitin ligase TRIM23, and induces K27-linked polyubiquitination of VDAC1, thereby promoting mitophagy and conferring robust neuroprotection. This TRIM23-VDAC1 axis represents a druggable target, as evidenced by the synergistic therapeutic efficacy of RSV and RGCs-targeted Trim23 overexpression in both diabetic and acute injury models (Fig. 9).

Fig. 9.

Resveratrol attenuates mitochondrial damage and protects RGCs by promoting TRIM23-mediated VDAC1 ubiquitination and mitophagy

Under HG stress, RSV binds to VDAC1 and recruits TRIM23, enabling K27-linked ubiquitination of VDAC1. This ubiquitination initiates the recognition and clearance of damaged mitochondria via the p62-LC3 autophagic pathway, thereby alleviating RGCs neurodegeneration.

VDAC1 has emerged as a central node in mitochondrial-dependent apoptosis across neurodegenerative contexts [10]. Our data firmly position it within the pathogenesis of DR. The consistent upregulation of VDAC1 in human donor retinas, diabetic mice, and HG-stressed neurons, coupled with the profound protection afforded by its knockdown, confirms its active role as a driver of RGCs loss, moving beyond a correlative biomarker. This aligns with the neuroprotective effect of VDAC1 inhibition in other injury models, suggesting a common pathogenic thread in neuronal degeneration [9, 21, 26]. Therefore, therapeutic strategies aimed at controlling VDAC1’s deleterious functions are warranted.

Impaired autophagic flux is recognized as a key contributor to RGCs degeneration in diabetes, as demonstrated by the protective role of the Sirt5-OPTN desuccinylation axis [27]. Here, we identify the natural polyphenol RSV as a therapeutic agent that can rectify this autophagic deficit through a novel, direct target. Contrary to the traditional view of RSV acting through indirect pathways like SIRT1, we show it directly binds to VDAC1 [28]. Notably, this interaction is also observed in cancer models, where RSV binding to VDAC1 promotes mitochondrial permeability transition and apoptosis, as well as in neurodegenerative contexts such as Parkinson’s disease, where RSV inhibits VDAC1-mediated mitochondrial dysfunction, reduces α-synuclein accumulation in mitochondria, stabilizes the mitochondrial permeability transition pore, and mitigates dopaminergic neuron degeneration and pathological progression [19, 21]. The stark contrast in functional outcomes underscores the context-dependence of VDAC1 modulation. Our mechanistic data clarify this paradox: we propose that in diabetic neurons, RSV binding acts as an allosteric switch, reconfiguring VDAC1 to favor TRIM23-mediated K27-ubiquitination and mitophagic clearance, thereby diverting its fate from the pro-apoptotic oligomerization observed in malignant cells.

The identification of TRIM23 as the responsible E3 ligase and the specific assignment of K27-linked ubiquitination are central to our model. This non-canonical ubiquitin topology typically serves as a signal for selective autophagy rather than proteasomal degradation. This elegantly explains the observed paradox: K27-ubiquitination does not mark VDAC1 for proteasomal decay but instead flags its host mitochondrion for autophagic clearance, consistent with the concomitant activation of LC3-II and p62 degradation. This TRIM23-VDAC1 pathway operates distinctly from the canonical PINK1-Parkin axis, revealing a new layer of regulation in mitochondrial quality control. Our finding that TRIM23, but not TRIM21, binds VDAC1 underscores the specificity within the TRIM family. This specificity is further highlighted by a recent study in nasopharyngeal carcinoma models, which revealed that TRIM21 specifically promotes the K48-linked ubiquitination and degradation of VDAC2, but does not bind to VDAC1, to regulate mitochondrial DNA release and antitumor immunity [29]. Notably, this is not an isolated case; a parallel can be drawn with a recent study demonstrating that TRIM31, another TRIM family member, binds and promotes the proteasomal degradation of VDAC1 to alleviate neurodegeneration in a Parkinson’s disease model [30]. Together, these findings solidify the emerging paradigm that distinct TRIM E3 ligases can fine-tune neuronal survival by targeting VDAC1 through different ubiquitin-dependent mechanisms.

The translational potential of the TRIM23-VDAC1 axis is strongly underscored by our combination therapy, in which RSV plus RGC-targeted Trim23 overexpression produced synergistic rescue of retinal structure and function in both chronic (STZ-induced DR) and acute (AOH) models, effectively transcending hyperglycemia-dependent and ischemia-driven insults. The lack of additive effect on retinal thickness in the STZ model may be due to a ceiling effect of RSV alone on structural preservation, or because thickness changes are slower to manifest than functional improvements. Nevertheless, the gains in ERG and RGCs counts support a neuroprotective synergy of the TRIM23-VDAC1 axis.

The robust neuroprotection observed in the rapid AOH paradigm, a well-validated model of isolated RGCs injury independent of systemic metabolic confounders, confirms the direct, cell-autonomous protective capacity of this pathway and positions it as a core component of the retina’s endogenous defense system. Given RSV’s established clinical safety profile, its emerging nanoformulations with enhanced ocular bioavailability, and the maturation of AAV2-SNCG vectors for safe, RGC-specific gene delivery, rapid clinical repositioning is highly feasible [31, 32]. Combining oral/intravitreal RSV with Trim23-overexpressing AAV or next-generation small-molecule TRIM23 activators could provide a powerful dual-hit strategy to halt RGCs loss in early DR, non-proliferative stages, and even acute glaucomatous or ischemic events.

While the combination of RSV and RGC-specific AAV2-SNCG-Trim23 delivery at 60 days post-STZ showed significant functional benefits over RSV alone in diabetic mice, we acknowledge that earlier TRIM23 induction might further enhance therapeutic efficacy. Previous work has shown that sustained AAV2 transduction is well-tolerated in RGCs. Zheng et al. demonstrated that rAAV2-transduced RGCs can be permanently transduced, with 97% of initially labeled cells still detectable at day 375 post-injection, supporting the long-term stability of AAV2-mediated expression in RGCs [33]. Future studies using preventive gene delivery paradigms or inducible systems will be valuable to further optimize the therapeutic window for this RGC-targeted approach.

Despite the robust mechanistic insights gained, this study has certain limitations that warrant consideration. A primary limitation lies in the use of the SH-SY5Y neuroblastoma cell line. The SH-SY5Y human neuroblastoma cell line is one of the most widely used in vitro neuronal models for studying neuroprotective mechanisms and signaling pathways. Although not derived from retinal tissue, SH-SY5Y cells have been extensively employed in RGC-related studies, including investigations of neuroserpin dysfunction in glaucoma [34], CD82-mediated axonal protection [35], diabetic retinal neurodegeneration [36]. Furthermore, the precise structural alterations in VDAC1 induced by RSV binding remain to be elucidated. Techniques such as cryo-electron microscopy could provide atomic-level details of this interaction, guiding the rational design of more potent and specific VDAC1-targeting therapeutics. Third, although we used an RGC-specific SNCG promoter to overexpress Trim23, we cannot completely exclude the possibility that systemic RSV may exert indirect protective effects via other retinal cell types. Future studies using RGC-specific Vdac1 knockout mice would help definitively establish the cell-autonomous requirement of VDAC1 in RSV-mediated neuroprotection.

In summary, our work progresses from target identification to mechanism elucidation and therapeutic validation. We establish a linear pathway from a small molecule binding its target to the execution of a protective cellular process, all orchestrated through a previously unrecognized TRIM23-VDAC1 ubiquitination cascade. This not only provides a new mechanistic framework for RSV’s action and mitochondrial quality control in the retina but also opens the door to innovative combination therapies designed to co-opt this natural protective system for the treatment of diabetic and other neurodegenerative retinopathies.

Conclusion

In summary, this study delineates a complete and previously unrecognized molecular pathway that underlies RSV’s neuroprotective action in DR. We have established that the mitochondrial gatekeeper protein VDAC1 is a pivotal pathogenic factor and a direct therapeutic target of RSV in RGCs. Moving beyond RSV’s known indirect effects, we demonstrated that its binding to VDAC1 initiates a precise molecular cascade: the allosteric recruitment of the E3 ubiquitin ligase TRIM23, which in turn catalyzes K27-linked polyubiquitination of VDAC1, thereby tagging the entire mitochondrion for clearance via mitophagy. The indispensability of this TRIM23-VDAC1 axis was confirmed by its necessity for RSV’s effects and, most importantly, by the profound synergistic neuroprotection achieved through its therapeutic augmentation in multiple disease models. Therapeutically, our findings position the resveratrol-TRIM23-VDAC1 axis as a promising target for developing combination therapies aimed at halting the progression of diabetic neurodegeneration.

Authors' contributions

R.H.W., X.F.S. and Y.Z. designed the experiments. R.H.W. performed the experiments. R.H.W., X.H.M., Z.Z. and L.Q.L. analyzed raw data. Y.X.M., H.M.Y., Y.H.T. and L.R. reviewed the data and made substantial contributions to improving the studies. R.H.W. wrote the manuscript.

Funding

This work was supported by the National Key Research and Development Program of China (Grant No. 2024YFC2510803) and the National Natural Science Foundation of China (Grant No. 82471103 and U24A20707).

Data availability

Data will be made available on request.

Declarations

Competing interests

The authors declare no competing interests.