Neddylation-Mediated hnRNPA2B1 Degradation Aggravates Retinal Endothelial Cell Dysfunction in Diabetic Retinopathy by Regulating miR-93-5p/VEGFA

Abstract

Purpose

Retinal endothelial cells (RECs) are key targets of diabetes-induced microvascular complications. HnRNPA2B1 suppresses pathological neovascularization in diabetic retinopathy (DR). Although hnRNPA2B1 suppresses pathological neovascularization, its role in hyperglycemia-induced REC dysfunction remains unclear.

Methods

Primary mouse retinal vascular endothelial cells (mRVECs) under high-glucose (HG) conditions and streptozotocin-induced diabetic mice were analyzed using quantitative real-time PCR (qRT-PCR), Western blotting, RNA immunoprecipitation, immunofluorescence staining, and functional assays (wound healing, Transwell invasion, and tube formation). Co-immunoprecipitation and pharmacological inhibitors were used to validate protein interactions and degradation pathways. Retinal morphology and vascular integrity were assessed using hematoxylin–eosin staining, optical coherence tomography angiography, Evans blue leakage, and trypsin digestion.

Results

HG-induced neddylation mediated hnRNPA2B1 degradation, exacerbating REC dysfunction. Mechanistically, hnRNPA2B1 facilitated miR-93-5p maturation by recruiting DGCR8 within the microprocessor complex, thereby suppressing VEGFA expression via direct targeting of its 3′-untranslated regions. Intravitreal delivery of AAV2-hnRNPA2B1 or miR-93-5p into diabetic mice partially restored retinal hnRNPA2B1/miR-93-5p levels, reduced VEGFA overexpression, and improved retinal histological markers of microvascular damage.

Conclusions

HG-induced effects associated with neddylation pathways lead to hnRNPA2B1 degradation, exacerbating REC dysfunction. HnRNPA2B1, as an RNA binding protein, facilitated miR-93-5p maturation by recruiting DGCR8 within the microprocessor complex. Targeting either hnRNPA2B1 or miR-93-5p may represent a potential therapeutic strategy for preserving retinal vascular homeostasis in diabetes pending functional validation.

Diabetic retinopathy (DR), a leading microvascular complication of diabetes, is projected to affect 160.5 million patients globally by 2045,¹ posing a severe healthcare burden. Retinal endothelial cells (RECs), the primary target of hyperglycemic damage, are critical for maintaining retinal homeostasis.^2,3 REC dysfunction leads to blood–retinal barrier (BRB) breakdown, resulting in pathological features, such as increased vascular permeability and acellular capillaries formation,⁴ ultimately causing irreversible vision loss. Early interventions targeting REC dysfunction have substantial clinical significance in delaying DR progression.

Heterogeneous nuclear ribonucleoprotein A2B1 (hnRNPA2B1), a member of the hnRNP family of RNA-binding proteins,⁵ coordinates RNA metabolic processes, including mRNA splicing,⁶ microRNAs (miRNA) processing,^7,8 and RNA exosome sorting.^9–11 Although hnRNPA2B1 suppresses pathological neovascularization via the STAT4/miR-223-3p/FBXW7 axis in DR,^12,13 its role in hyperglycemia-induced REC dysfunction remains unclear.

The miRNAs represent a class of short non-coding RNAs that modulate gene expression by binding to the 3′-untranslated region (3′-UTR) of target mRNAs. Through this post-transcriptional mechanism,^14–16 miRNAs are pivotal regulators of diverse biological processes.¹⁷ In DR, protective miRNAs, such as miR-29b-3p¹⁸ and miR-152,¹⁹ attenuate DR progression by suppressing VEGF expression, whereas pathogenic miRNAs, such as miR-183, exacerbate VEGF-mediated angiogenesis via BTG1 downregulation.²⁰ Conversely, reduced miR-590-3p levels in DR promote endothelial pyroptosis via NLRP1 activation and IL-1β/NOX4-dependent pathways, ultimately suppressing neovascularization.²¹

In the current study, high-glucose (HG)-induced, neddylation-mediated hnRNPA2B1 degradation and its role in diabetic retinal microvascular dysfunction were investigated. Using mouse primary retinal vascular endothelial cells (mRVECs) under HG conditions and streptozotocin (STZ)-induced diabetic models, this study demonstrated that hnRNPA2B1 alleviated diabetes-associated retinal dysfunction and microvascular damage by promoting miR-93-5p maturation, thereby suppressing the upregulation of pathogenic VEGFA. Retinal miR-93-5p overexpression substantially attenuated DR progression, highlighting its therapeutic potential in diabetic microangiopathy.

Methods

Cell Culture

Primary mRVECs were procured from ProCell (CP-M114, Wuhan, China), maintained in basal endothelial cell medium fortified with 10% fetal bovine serum (FBS) under standard culture conditions (37°C, 5% CO₂), and designated as the normal group. For HG treatment, cells were exposed to 30 mM glucose (Solarbio, Beijing, China) in parallel culture. Cells were treated with 5 mM glucose (CON), 30 mM glucose (HG), or 5 mM glucose + 25 mM mannitol (OSM) for 48 hours before analysis.

Lentivirus Transfection of hnRNPA2B1 in mRVECs

The hnRNPA2B1 overexpression construct was engineered by GeneChem (Shanghai, China) by inserting the target gene into the GV146 plasmid (CMV-MCS-IRES-EGFP-SV40-Neomycin). Knockdown vectors encoding shRNAs targeting hnRNPA2B1 were cloned into the GV102 plasmid (hU6-MCS-CMV-GFP-SV40-Neomycin). Lentiviral particles were generated through transient transfection of HEK293T cells, followed by the collection of viral supernatants for subsequent in vivo applications. Stable transfectants were selected using neomycin- or puromycin-resistant markers.

Diabetic Retinopathy Model

All experimental protocols were endorsed by the Experimental Animal Administration Committee of Qilu Hospital, Shandong University (approval no.: DWLL-2024-193). Male C57BL/6J mice (Beijing Vital River Laboratory Animal Technology Co., Ltd., Beijing, China), maintained under specific pathogen-free conditions (22°C ± 2°C, 50%–60% humidity, 12-hour light/dark cycle) with as desired access to food and water, underwent a seven-day acclimatization period before the induction of diabetes. After overnight fasting, type 1 diabetes was established by the intraperitoneal administration of STZ (50 mg/kg; Sigma-Aldrich Corp., St. Louis, MO, USA) prepared in citrate buffer on five consecutive days. Control cohorts received citrate buffer alone. Diabetic status was confirmed one week after STZ injection using a blood glucose meter to measure blood glucose concentrations exceeding 16.7 mmol/L over seven consecutive days. Body weight and glycemic parameters were monitored biweekly.

Intravitreal Injection of AAV2 Vector

After eight weeks after diabetes induction, mice were randomized to AAV treatment groups using number tables, with no baseline differences in blood glucose or body weight. Mice were anesthetized using tribromoethanol (30 µL/g; Meilunbio) and the right eye dilated with 1% atropine sulfate and 2.5% phenylephrine hydrochloride. Adeno-associated serotype 2 virus (AAV2) vectors (GeneChem) were produced via triple-plasmid co-transfection in HEK293T cells, purified by ultracentrifugation, and titered by qPCR. Viral suspension (1.5 µL; AAV2-hnRNPA2B1: 4.52 × 10¹² vg/mL, AAV2-miR-93-5p: 4.24 × 10¹² vg/mL) was intravitreally injected (delivering ∼6.78 × 10⁹ and 6.36 × 10⁹ vg/eye, respectively). The dose was selected based on literature (10⁹–10¹¹ vg/eye) and preliminary efficacy. Over four weeks after injection, regular slit-lamp examinations revealed no overt inflammation, vitreous opacities, or retinal toxicity, and no systemic adverse effects were observed.

QRT-PCR Analysis

RNA was extracted from the retinal tissue and mRVECs were treated with TRIzol reagent (Hunan Accurate Bio-Medical Co., Ltd., Hunan, China) following standard procedures. Reverse transcription was carried out using the PrimeScript RT Reagent Kit (Toyobo Co., Ltd., Osaka, Japan) to generate cDNA templates. Quantitative PCR amplification was conducted using the SYBR Premix Ex Taq with GAPDH as the endogenous reference for mRNA and U6 snRNA for miRNA normalization. Relative expression levels were determined using comparative Cq (2^−ΔΔCq) analysis. For the analysis of mRNA decay, ActD (HY-17559; MedChemExpress, Monmouth Junction, NJ, USA) was introduced to the medium for specified durations. Primer sequences are provided in Supplementary Material ‘primer list.docx’.

Western Blotting

Protein lysates were prepared with cell lysis buffer (P0013; Beyotime Biotechnology, Shanghai, China). For subcellular fractionation, nuclear/cytoplasmic extraction kits (P0027; Beyotime Biotechnology) with protease/phosphatase inhibitor cocktails (P1050; Beyotime Biotechnology) were used. After BCA quantification, 10 µg/µL samples were separated on 10% SDS-PAGE gels and transferred to PVDF membranes (Millipore, Burlington, MA, USA). Membranes were blocked in 5% nonfat milk for three hours, then incubated overnight at 4°C with primary antibodies: hnRNPA2B1 (83773-7-RR; Proteintech, Wuhan, China), NEDD8 (ab81264; Abcam, Cambridgeshire, UK), VEGFA (81323-2-RR; Proteintech), DGCR8 (10996-1-AP; Proteintech), and tubulin (11224-1-AP; Proteintech). After secondary antibody incubation (37°C, one hour) and Tris-buffered saline with Tween-20 washes, detection was performed using BeyoECL Star (Beyotime Biotechnology) with an Amersham Imager 600 RGB (Amersham, Piscataway, NJ, USA).

RNA Immunoprecipitation (RIP) Assay

The RIP assay was carried out using the EZMagna RIP kit (Sigma-Aldrich). Cell lysates were subjected to immunoprecipitation with anti-hnRNPA2B1 (14813-1-AP; Proteintech) or control IgG (Proteintech) at 4°C. RT-PCR was used to quantify the precipitated RNA.

Wound Healing Assay

Confluent mRVEC monolayers in 24-well plates were subjected to linear wounding using a sterile pipette tip. Cell migration was imaged at 0 and 24 hours under a phase-contrast microscope (Leica, Wetzlar, Germany).

Transwell Assay

Cells in serum-free medium were plated onto the upper chambers (Corning, Corning, NY, USA), with 10% FBS serving as a chemotactic stimulus in the lower compartments. After a 24-hour incubation, migrated cells were fixed with 4% paraformaldehyde, stained with 0.1% crystal violet, and quantified using image analysis software.

Tube Formation Assay

Matrigel-coated 24-well plates were incubated (37°C, one hour) before seeding mRVECs (1 × 10⁵ well). Capillary-like structures were imaged (Nikon, Melville, NY, USA) and analyzed for branch points after 24 hours.

Dual Luciferase Reporter Assay

Dual-luciferase reporter vectors harboring wild-type/mutant VEGFA 3′-UTR sequences (GeneChem) were co-transfected into HEK293T cells alongside miR-93-5p mimics or control oligonucleotides using Lipofectamine 3000 transfection reagent (Thermo Fisher Scientific, Waltham, MA, USA). Luciferase activity was assessed 48 hours after transfection (Yeasen Biotechnology, Shanghai, China).

Co-Immunoprecipitation

Lysates were immunoprecipitated with specific antibodies or IgG (A7016, Beyotime) overnight at 4°C. Protein A/G magnetic beads were then added for four hours at 4 °C, washed with Tris buffered saline solution, and eluted with loading buffer. The eluates were analyzed using immunoblotting.

Optical Coherence Tomography Angiography (OCTA)

The retinal vasculature was assessed with abdominal anesthesia and using OCTA with a small Animal Ophthalmology Multimodal Imaging System (Nanjing Boshi Medical Technology Co., Nanjing, China). The optical microangiography (OMAG) and 3D super-resolution algorithms generated vascular maps from the five B-scan composites. The vessel area density was computed as the proportion of the perfused vascular area to the total retinal area.

Hematoxylin and Eosin (H&E) Staining of Retinal Tissue

Retinas fixed with paraformaldehyde were embedded in paraffin, sectioned into 5-µm-thick slices, and subjected to HE staining. Morphological analysis was performed using a slide scanner (Olympus VS200; Olympus, Tokyo, Japan).

Retinal Trypsin Digestion Assay

Retinas were digested (3% trypsin, 37°C, three hours), rinsed, and stained with periodic acid–Schiff/hematoxylin. Vascular networks were imaged as previously described. Images were captured using a digital section-scanning system (Olympus VS200).

Evans Blue (EB) Assay

After intraperitoneal administration of tribromoethanol (30 µL/g; Dalian Meilun Biotech Co., Ltd., Dalian, China) for anesthesia, EB (20 mg/mL in saline solution; Solarbio) was delivered intravenously via the tail vein at 45 mg/kg. After two hours of systemic circulation, mice were euthanized by anesthetic overdose and transcardially perfused with PBS to clear intravascular tracer. Eyes were enucleated, retinas dissected, dried, and weighed. EB was extracted from dried retinas in formamide (120 µL, 70 °C, 18 h), and concentration was determined spectrophotometrically (absorbance 620 nm, background correction 740 nm) against a standard curve. Data are expressed as ng EB per mg retinal dry weight (ng/mg). For spatial visualization, retinal whole-mount preparations were imaged using a Zeiss LSM880 confocal system (Zeiss, Oberkochen, Germany) to detect EB leakage.

Immunofluorescence

Cell slides and sections were permeabilized with 0.2% Triton X-100 for 30 minutes. Sections were then blocked with 5% FBS for two hours and incubated with antibodies at 4°C overnight. Fluorescently-conjugated secondary antibodies were applied, followed by DAPI counterstaining before imaging (Olympus VS200; Olympus).

Blinded Assessment

Image acquisition (OCTA, fluorescence microscopy) and quantitative analysis were performed by researchers blinded to group allocation.

Statistical Analyses

Data are presented as the mean ± SD. All datasets were assessed for normality (Shapiro-Wilk test) and homogeneity of variance (Brown-Forsythe test). Statistical analyses were performed using GraphPad Prism 9.0 (GraphPad, San Diego, CA, USA). Comparisons between two groups were conducted using two-tailed unpaired Student's t tests. For comparisons among multiple groups, one-way ANOVA was applied, followed by post hoc multiple comparisons with Tukey's or Bonferroni's correction where indicated. Statistical significance was defined as P < 0.05. Significance levels are denoted by asterisks as *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; "ns" indicates not significant. No repeated or longitudinal measurements requiring mixed-effects models were included. For in vitro experiments, n represents biologically independent experiments; for in vivo studies, n represents individual mice.

Results

HnRNPA2B1 Affects MRVECs Function Under HG Conditions

To investigate the role of hnRNPA2B1 in DR progression, we first confirmed the metabolic specificity of our model. Osmotic controls (OSM) showed no significant difference from control (CON) in hnRNPA2B1 protein levels (Fig. 1A), cell migration (Fig. 1B), invasion (Fig. 1C), tube formation (Fig. 1D), and cell morphology (Fig. 1E), demonstrating that hyperosmolar stress alone does not affect baseline endothelial function. We next constructed controls, hnRNPA2B1 short-hairpin RNA (shRNA), and hnRNPA2B1 lentiviruses. Western blotting confirmed the successful suppression or upregulation of hnRNPA2B1 expression in mRVECs transfected with the corresponding vectors or shRNAs (Figs. 1F, 1G). Functional assays, including wound healing, Transwell, and tube formation assays, indicated that hnRNPA2B1 overexpression reversed the HG-induced migration, invasion, and angiogenesis of mRVECs. In contrast, hnRNPA2B1 knockdown exacerbated these pathological phenotypes (Figs. 1H–J). Taken together, hnRNPA2B1 substantially alleviated HG-induced dysfunction in mRVECs, highlighting its protective role in DR pathogenesis.

Figure 1.

The protective role of hnRNPA2B1 in mRVECs exposed to HG conditions. (A) OSM showed no significant difference in hnRNPA2B1 protein levels compared to CON. (B–E) Functional assays including wound healing (B), Transwell invasion (C), tube formation (D), and cell morphology (E) demonstrated that OSM treatment alone produced no significant effects compared to CON. (F, G) Western blotting confirmed successful upregulation (F) or suppression (G) of hnRNPA2B1 expression using lentivirus vectors. (H–J) Functional assays after hnRNPA2B1 manipulation showed that overexpression reversed HG-induced migration, invasion, and angiogenesis, whereas knockdown exacerbated these phenotypes. Quantitative data are from n = 3 biologically independent experiments. Bars: mean ± SD. Statistical analysis was performed by one-way ANOVA followed by Tukey's multiple comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

HnRNPA2B1 Alleviates Diabetes-Induced Retinal Dysfunction

The retinal protective effects of hnRNPA2B1 were assessed in vivo using STZ-induced diabetic mice. To achieve retina-specific hnRNPA2B1 overexpression, mice were intravitreally injected with AAV2 vectors. Western blotting (Fig. 2A) indicated that hnRNPA2B1 protein levels were substantially downregulated in diabetic retinas at 12 weeks after STZ treatment and were rescued by AAV2–hnRNPA2B1 transduction but not with AAV2–NC. Double immunofluorescence (IF) staining for hnRNPA2B1 (pink) and the endothelial marker, CD31 (red), confirmed these findings (Fig. 2B).

Figure 2.

hnRNPA2B1 attenuates hyperglycemia-induced retinal vascular dysfunction in vivo. (A) Western blotting was conducted to detect the protein levels of hnRNPA2B1 in the retinal tissue of mice at the twelfth week after STZ. (B) IF staining of hnRNPA2B1 (pink), CD31 (red), and DAPI staining (blue) of retinal tissue sections. Scale bar: 100 µm. (C, H) Retinal thickness was measured in HE-stained sections. (D, I) Retinal vessel area density was measured in OCTA. (E, J) Fluorescent signal quantification in retinal flat mounts was conducted via confocal immunofluorescence microscopy, with quantitative mapping of EB tracer distribution systematically demonstrating vascular hyperpermeability. (F) IF staining of ZO-1 (green), CD31 (red), and DAPI staining (blue) of retinal tissue sections. Scale bar: 100 µm. (G, K) PAS staining of retinal tissue quantified E/P ratio and acellular capillary density. Scale bar: 100 µm. Quantitative data (C–E, G–K) are from n = 6 biologically independent mice per group. Bars: mean ± SD. Normality was confirmed by Shapiro-Wilk test and equal variances by Brown-Forsythe test before performing one-way ANOVA with Tukey's post-hoc correction. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

Retinal structural integrity and functional morphology were evaluated. H&E staining (Figs. 2C, 2H) and OCTA (Figs. 2D, 2I) indicated characteristic DR features, including decreased retinal thickness and vascular density. The hnRNPA2B1 overexpression attenuated these pathological alterations in AAV2–NC-treated mice. DR triggered the disruption of the retinal microvasculature, leading to BRB hyperpermeability and pathological neovascularization. EB dye leakage assays demonstrated that vascular permeability in diabetic retinas increased markedly, which was partially restored by AAV2–hnRNPA2B1 intervention (Figs. 2E, 2J). Double IF staining for tight junction proteins, ZO-1 (green) and CD31 (red), confirmed that AAV2–hnRNPA2B1 treatment upregulated ZO-1 expression (Fig. 2F). The retinal trypsin digestion assay (Figs. 2G, 2K) indicated that hnRNPA2B1 overexpression effectively attenuated the diabetes-associated pericyte loss and acellular capillary formation. Collectively, these findings demonstrate that hnRNPA2B1 improves histological markers of diabetes-induced retinal dysfunction by preserving vascular integrity, restoring BRB function, and reducing pathological microvascular remodeling.

HG Suppresses MiR-93-5p Maturation by Downregulating HnRNPA2B1

VEGFA, a master regulator of angiogenesis and vascular permeability, exhibited stable RNA levels but markedly reduced protein expression after hnRNPA2B1 overexpression (Figs. 3A, 3B), suggesting the involvement of post-transcriptional regulation. Importantly, OSM showed no significant difference in VEGFA mRNA or protein levels compared to CON (Fig. 3C, 3D), confirming metabolic specificity. Triple IF staining for hnRNPA2B1 (pink), VEGFA (green), and CD31 (red) in mouse retinas confirmed these findings, demonstrating reduced VEGFA protein co-localization with the retinal vasculature in hnRNPA2B1-overexpressing tissues (Fig. 3E). The direct association of hnRNPA2B1 with DGCR8, an essential subunit of the nuclear miRNA microprocessor complex, facilitates the m⁶A-dependent processing of primary miRNA transcripts (pri-miRNAs).⁷ IF staining and nucleocytoplasmic fractionation indicated that hnRNPA2B1 was localized in the nucleus, consistent with its role in nuclear miRNA biogenesis (Figs. 3F, 3G). Confocal imaging further revealed the co-localization of both hnRNPA2B1 and DGCR8 in nucleus under normal glucose conditions. Although HG treatment significantly reduced the protein levels of both proteins, their co-localization pattern was maintained (Fig. 3H). This suggests that HG likely impairs microprocessor function by triggering protein degradation rather than by disrupting spatial assembly.

Figure 3.

HG inhibits miR-93-5p processing by suppressing hnRNPA2B1 expression. (A) VEGFA mRNA level was detected using qRT-PCR assays. (B) VEGFA protein level was detected using Western blotting. (C) VEGFA mRNA level under OSM conditions. (D) VEGFA protein level under OSM conditions. (E) IF staining of hnRNPA2B1 (pink), VEGFA (green), CD31 (red), and DAPI staining (blue) of retinal tissue sections. n = 6. Scale bar: 100 µm. (F, G) IF staining and Western blotting detected hnRNPA2B1 location in mRVECs. (H) IF staining detected hnRNPA2B1 (red) and DGCR8 (green) location in HG-treated (30 mM) or untreated (CON) mRVECs in mRVECs. (I) Co-IP of DGCR8 and hnRNPA2B1 under pretreatment with RNaseA/T in mRVECs. (J) QRT-PCR detected the expression of miRNAs in HG-treated (30 mM) or untreated (CON) mRVECs. (K) qRT-PCR detected pri-miR-93, pre-miR-93 and mature miR-93-5p levels under OSM conditions. (L) qRT-PCR detected the expression of pri-miR-93, pre-miR-93 and mature miR-93-5p after hnRNPA2B1 overexpression. (M) DROSHA and DGCR8 protein level was detected using Western blotting. (N) HnRNPA2B1 and DGCR8 protein level was detected using Western blotting. (O) QRT-PCR detected the expression of pri-miR-93, pre-miR-93 and mature miR-93-5p after hnRNPA2B1 overexpression or DGCR8 downexpression. (P) RIP assay analyzed the binding of hnRNPA2B1 with pri-miR-93-5p after hnRNPA2B1 overexpression. Quantitative data (A, B, H–M) are from n = 3 biologically independent experiments. Bars represent mean ± SD. Statistical analysis was performed by one-way ANOVA with Tukey's test for multiple comparisons. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

Co-IP assays validated the endogenous interaction between hnRNPA2B1 and DGCR8 in mRVECs under pre-treatment of RNaseA/T (Fig. 3I). Based on these findings, we hypothesized that hnRNPA2B1 regulated VEGFA expression via miRNA-mediated posttranscriptional silencing.

To investigate the differentially expressed miRNAs in diabetic retinas, the GSE160310 dataset was analyzed and 238 significantly downregulated miRNAs were identified (|logFC| > 0.2, P < 0.05). Using TargetScan-based target mRNA prediction and Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis, 10 angiogenesis-associated miRNAs were screened. Subsequent qRT-PCR assays indicated a significant downregulation of mmu-miR-93-5p in mRVECs under HG conditions (Fig. 3J).

To investigate whether hnRNPA2B1 regulated maturation, the precursor and mature miRNA levels were analyzed. OSM for pri/pre/mature miRNA levels showed no significant difference from CON (Fig. 3K). The hnRNPA2B1 overexpression reduced pri-miR-93 levels while increasing pre-miR-93 and mature miR-93-5p levels (Fig. 3L). HG also lowered Drosha and DGCR8 proteins, while osmotic controls showed no effect (Fig. 3M); hnRNPA2B1 reinstated DGCR8 levels (Fig. 3N). si-DGCR8 abolished the hnRNPA2B1-mediated increase in mature miR-93-5p (Fig. 3O), confirming that hnRNPA2B1 promotes miR-93-5p maturation through DGCR8-dependent microprocessor assembly. RIP assays further demonstrated that the hnRNPA2B1 antibody resulted in greater pri-miR-93 enrichment compared with the IgG controls, with enhanced binding efficiency under hnRNPA2B1 overexpression (Fig. 3P). These findings suggest that hnRNPA2B1 promotes miR-93-5p maturation by facilitating microprocessor complex assembly.

MiR-93-5p Regulates MRVEC Function by Targeting VEGFA

The mitigating effects of hnRNPA2B1 overexpression on HG-induced migration, invasion, and tube formation of mRVECs were removed by miR-93-5p inhibition (Figs. 4A–C). These results suggest that hnRNPA2B1 regulated mRVEC function via a miR-93-5p dependent mechanism. Building on this evidence and the established function of miR-93-5p in diabetes, hnRNPA2B1 was thought to attenuate HG-induced mRVEC dysfunction through miR-93-5p–VEGFA signaling. TargetScan analysis indicated a conserved miR-93-5p binding motif within the VEGFA mRNA 3′-UTR (Fig. 4D), thereby supporting direct posttranscriptional regulation.

Figure 4.

hnRNPA2b1 regulates mRVEC function through targeting miR-93-5p/VEGFA under HG conditions. (A) The migration ability of mRVECs was detected using the wound healing assay. (B) The invasion ability of mRVECs was assessed using the transwell assay. (C) The angiogenesis ability of cells was evaluated using the tube formation experiment. (D) TargetScan predicted the binding sites between miR-93-5p and VEGFA 3′UTR. (E) QRT-PCR detected the expression of miR-93-5p after transfection with mimics control, miR-93-5p mimics, inhibitor control, or miR-93-5p inhibitor in HEK293T cells. (F) The luciferase activity was quantified and normalized relative to that of the expressed Renilla luciferase activity. (G) VEGFA protein level was detected using Western blotting. Quantitative data (A–C and F) are from n = 3 biologically independent experiments. Two-group comparisons (E and G) are from n = 3 independent experiments. Bars represent mean ± SD. One-way ANOVA with Tukey's test was applied for A–C and F. Two-tailed unpaired t-test was used for E and G. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

Subsequently, HEK293T cells were transfected with wild type (WT)/mutant (Mut) VEGFA-3′-UTR constructs and miR-93-5p mimics, inhibitors, or controls. The qRT-PCR assays confirmed that mimics upregulated, whereas inhibitors downregulated miR-93-5p in HEK293T cells (Fig. 4E). Dual-luciferase reporter assays demonstrated that WT-VEGFA-3′-UTR luciferase activity was suppressed by miR-93-5p mimics, whereas this was enhanced by inhibitors. The Mut-VEGFA–3′-UTR construct had no marked effect (Fig. 4F). Western blotting assays indicated that miR-93-5p inhibition in hnRNPA2B1-overexpressing mRVECs under HG conditions increased VEGFA protein levels, whereas miR-93-5p mimics reduced VEGFA expression (Fig. 4G). Collectively, the findings indicated that hnRNPA2B1 suppresses VEGFA expression by promoting the maturation of miR-93-5p, which directly silences VEGFA as a direct posttranscriptional repressor.

MiR-93-5p Alleviates Diabetes-Induced Retinal Dysfunction

To evaluate the therapeutic potential of miR-93 in vivo, AAV2–miR-93-5p or AAV2–NC were intravitreally injected into mice. The qRT-PCR assays indicated a reduction in retinal miR-93-5p levels 12 w after STZ induction, which was restored following AAV2–miR-93-5p treatment (Fig. 5A). Western blotting indicated that the upregulation of miR-93-5p mitigated the diabetes-induced overexpression of VEGFA (Fig. 5B). Double IF staining for VEGFA (green) and CD31 (red) confirmed a marked decrease in VEGFA protein levels within retinal vascular endothelial cells after AAV2–miR-93-5p intervention (Fig. 5C).

Figure 5.

MiR-93-5p improves histological markers of diabetes mellitus-induced retinal vascular dysfunction in vivo. (A) QRT-PCR detected the expression of miR-93-5p. (B) Western blotting detected VEGFA protein levels in the retinal tissues of mice at the twelfth week. (C) IF staining of VEGFA (green), CD31 (red), and DAPI staining (blue) of retinal tissue sections. Scale bar: 100 µm. (D, I) Retinal thickness was measured in HE-stained sections. (E, J) Retinal vessel area density was measured in OCTA. (F, K) Fluorescent signal quantification in retinal flat mounts was conducted via confocal immunofluorescence microscopy, with quantitative mapping of EB tracer distribution systematically demonstrating vascular hyperpermeability. (G) IF staining of ZO-1 (green), CD31 (red), and DAPI staining (blue) of retinal tissue sections. Scale bar: 100 µm. (H, L) PAS staining of retinal tissue quantified E/P ratio and acellular capillary density. Scale bar: 100 µm. Quantitative data (A, B, D–F, H–L) are from n = 6 biologically independent mice. Bars: mean ± SD. Statistical analysis was performed by one-way ANOVA with Tukey's post-hoc correction. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

Next, retinal structural and functional integrity were evaluated. H&E staining and OCTA indicated that miR-93-5p overexpression ameliorated pathological retinal alterations (including reduced thickness and vascular density) in diabetic mice, compared to the AAV2–NC-treated controls (Figs. 5D, 5E, 5I, 5J). EB leakage assays indicated that AAV2–miR-93-5p intervention alleviated diabetes-induced blood–retinal barrier hyperpermeability (Figs. 5F, 5K). Double IF staining for ZO-1 (green) and CD31 (red) confirmed that AAV2–miR-93-5p treatment upregulated ZO-1 expression (Fig. 5G). Retinal trypsin digestion assays indicated that the upregulation of miR-93-5p reduced pericyte loss and acellular capillary formation in diabetic mice (Figs. 5H, 5L). Together, these findings demonstrate that modulating the hnRNPA2B1/miR-93-5p/VEGFA pathway through miR-93-5p upregulation suppressed VEGFA protein synthesis, thereby improving histological markers of diabetic retinal damage and microvascular pathology.

HG Promotes Neddylation-Mediated HnRNPA2B1 Protein Degradation in MRVECs

To investigate the role of hnRNPA2B1 in DR, mRNA and protein levels in HG-exposed mRVECs were investigated. A glucose concentration-dependent discordance was observed: hnRNPA2B1 mRNA levels increased, whereas its protein level decreased (Figs. 6A, 6B). First, mRNA stability assays were conducted using actinomycin D (ActD) to inhibit mRNA transcription; HG increased hnRNPA2B1 mRNA stability (Fig. 6C). The hnRNPA2B1 protein downregulation occurred concomitantly with the increased expression of the ubiquitin-like protein, NEDD8 (Fig. 6A). To delineate the involvement of ubiquitination or neddylation in hnRNPA2B1 regulation under HG stress, mRVECs were treated with the neddylation inhibitor, MLN4924; ubiquitination inhibitor, MLN7243; or proteasome inhibitor, MG132, under HG exposure. Western blotting indicated that MLN4924 effectively restored hnRNPA2B1 protein levels, whereas the other compounds exhibited no marked effect (Fig. 6D). Importantly, OSM showed no significant effect on NEDD8 protein expression, confirming that the metabolic changes observed under HG were specific rather than due to hyperosmolar stress (Fig. 6E). To genetically confirm the requirement of neddylation, siRNA-mediated knockdown of NEDD8 restored hnRNPA2B1 protein levels under HG (Fig. 6F). Computational docking analysis suggested a potential binding interface between hnRNPA2B1 (PDB: 5WWE) and NEDD8 (PDB: 1NDD) (Fig. 6G), and immunofluorescence revealed nuclear co-localization of both proteins that was diminished by HG but preserved by MLN4924 (Fig. 6H). Co-IP assays showed that the NEDD8–hnRNPA2B1 association was disrupted by MLN4924, and unaffected by MLN7243 and MG132 (Figs. 6I–K), supporting the hypothesis that hnRNPA2B1 is a neddylation substrate. In summary, these findings support a model where HG treatment increased hnRNPA2B1 mRNA stability while promoting its proteasomal degradation associated with the neddylation pathway.

Figure 6.

The degradation of hnRNPA2B1 protein is mediated by neddylation in mRVECs. mRVECs were treated with glucose at the indicated concentrations for 48 hours. (A) HnRNPA2B1 and NEDD8 protein levels were detected using Western blotting assays. (B) HnRNPA2B1 mRNA levels were detected using qRT-PCR assays. (C) QRT-PCR was conducted to detect the mRNA levels of hnRNPA2B1 in mRVECs after ActD treatment. (D) MRVECs were cultured under 5 mM or 30 mM glucose for 48 hours, and cells under 30 mM glucose was treated with 5 µM MLN4924, 1.5 µM MLN7243, or 1.5 µM MG132 for 48 hours. Western blotting was conducted to detect the protein levels of hnRNPA2B1. (E) NEDD8 protein expression under CON, HG, and OSM conditions was assessed by Western blotting. (F) SiRNA-mediated NEDD8 knockdown efficiency and its effect on hnRNPA2B1 protein level under HG conditions were assessed by Western blotting. (G) Molecular docking of hnRNPA2B1 (PDB: 5WWE) and NEDD8 (PDB: 1NDD) was performed using HDOCK server and PyMOL. (H) IF staining detected hnRNPA2B1 (red) and NEDD8 (green) localization in mRVECs under untreated (CON), HG-treated (30 mM), or HG + MLN4924 conditions. (I) Co-IP of NEDD8 and hnRNPA2B1 under HG with MLN7243 treatment. (J) Co-IP of NEDD8 and hnRNPA2B1 under HG with MG132 treatment. (K) Co-IP of NEDD8 and hnRNPA2B1 under HG with MLN4924 treatment. For Western blots and Co-IP images, blots are representative of n = 3 biologically independent experiments. Quantified data (A–E) are from n = 3 biologically independent experiments with technical triplicates. Bars: mean ± SD. Statistical analysis was performed by one-way ANOVA with Tukey's test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

Discussion

Diabetic retinopathy, a microvascular complication of diabetes, is characterized by early pathological alterations, including BRB dysfunction and vascular leakage. RECs, which are critical components of the BRB,²² are fundamentally involved in the pathophysiological progression of DR. Our research demonstrated that HG-induced neddylation enhanced the proteasomal degradation of hnRNPA2B1, which mechanistically regulated retinal endothelial dysfunction by regulating miR-93-5p maturation and elevated VEGFA expression (Fig. 7). Our findings identify hnRNPA2B1 as a candidate therapeutic target for DR that requires further functional validation.

Figure 7.

Neddylation-mediated hnRNPA2B1 degradation aggravates retinal endothelial cell dysfunction.

The hnRNPA2B1 has previously been demonstrated to have a role in vascular protection, including inhibition of endothelial proliferation and angiogenesis in coronary artery disease²³ and modulation of NF-κB-mediated anti-inflammatory responses.²⁴ In the current study, hnRNPA2B1 overexpression counteracted the HG-induced pathological activation of RECs, substantially attenuating migration, invasion, and tube formation. Moreover, overexpression of retinal hnRNPA2B1 in diabetic mice substantially restored retinal integrity and reduced microvascular damage. While neddylation-mediated degradation of Nrf2 exacerbated inflammatory and oxidative stress states in DR,²⁵ the current study indicated that HG upregulated NEDD8 expression in mRVECs, driving hnRNPA2B1 degradation via neddylation, but not ubiquitination, as validated using pharmacological inhibitor and co-IP assays.

The miRNA-mediated posttranscriptional regulation is critical for DR pathogenesis. The miRNA biogenesis is initiated via the processing of pri-miRNAs within the nuclear microprocessor machinery.²⁶ The hnRNPA2B1, as a RNA-binding protein, participates in m6A-dependent miRNA biogenesis by interacting with DROSHA and DGCR8.⁷ The current study demonstrated that hnRNPA2B1 promoted miR-93-5p maturation through interaction with DGCR8. While hyperosmolarity has been implicated in diabetic retinopathy through TonEBP/COX-2 signaling pathways,²⁷ our osmotic control experiments demonstrated that the observed hnRNPA2B1 degradation and miR-93-5p dysregulation were specific to metabolic glucose sensing rather than osmotic stress. Clinical profiling of patients with DR indicated significant downregulation of miR-93-5p in vitreous humor samples,²⁸ with its protective role extending to blood–brain barrier dysfunction via VEGFA suppression.²⁹ Notably, miR-93-5p exerts protective effects in DR by targeting Sirt1,³⁰ suggesting additional mechanisms of action, while the hnRNPA2B1/miR-93-5p axis has also been implicated in other diseases such as prostate cancer, highlighting its broad functional relevance.³¹ VEGFA exhibits potent angiogenic activity and increases vascular permeability, leading to enhanced vascular leakage.³² Although VEGFA is critical for retinal development,^33,34 its dysfunction triggers neovascular eye diseases.^35,36 By integrating bioinformatics analysis of the GSE160310 dataset with functional validation, miR-93-5p was mechanistically defined as a direct posttranscriptional silencer of VEGFA through 3′-UTR targeting under HG conditions.

The therapeutic modulation of miRNAs, such as mimic delivery or inhibitor administration, represents a potential therapeutic strategy for DR.^37,38 In the current study, AAV2-mediated intravitreal delivery of miR-93-5p was used to achieve sustained expression.³⁹ The miR-93-5p overexpression protected against diabetes-induced retinal dysfunction and microvascular pathology. Consistent with this, Western blotting and IF staining confirmed VEGFA downregulation both in vivo and in vitro, directly linking miR-93-5p protective effects to VEGFA suppression. Notably, the miR-93-5p binding site in VEGFA 3′UTR is evolutionarily conserved, suggesting fundamental physiological significance in vascular homeostasis. In DR, hyperglycemia-induced hnRNPA2B1 degradation impairs miR-93-5p maturation, leading to functional inactivation of this conserved angiogenic suppression module and derepression of VEGFA translation. This suggests how an evolutionarily conserved miRNA regulatory network becomes pathologically inactivated in metabolic disease. These preclinical mechanistic findings require further functional validation in visual performance assays (e.g., ERG or optokinetic behavior) before clinical translation.

Conclusions

This study establishes HG-induced neddylation as the critical upstream regulator destabilizing hnRNPA2B1, thereby impairing miR-93-5p maturation and driving VEGFA overexpression. Targeting this axis—either through hnRNPA2B1 stabilization or miR-93-5p delivery—offers a dual therapeutic strategy for DR. Recent advances in biomaterial platforms further validate this approach: hyaluronic acid hydrogels enable sustained miRNA release over 45 days,⁴⁰ and co-delivery of miR-21-3p antagomir with aflibercept circumvents conventional intravitreal therapy's constraints (frequent injections, VEGF neurotoxicity) while conferring synergistic vascular and neuroprotective effects.⁴¹ Likewise, neddylation inhibitors such as MLN4924, though currently evaluated in oncologic trials,⁴² provide a compelling precedent for stabilizing proteins like hnRNPA2B1.

A limitation of the mechanistic characterization is that while our data support neddylation-mediated hnRNPA2B1 degradation via the proteasome, alternative degradation pathways such as autophagy-lysosome were not systematically examined. Additionally, definitive biochemical proof would be required to fully establish hnRNPA2B1 as a direct NEDD8 substrate. Notably, the absence of functional retinal readouts such as ERG or visual behavior assays, in this study limits translational interpretation of the AAV-mediated rescue. Future studies should incorporate these functional endpoints to validate therapeutic potential and directly assess these strategies to attenuate DR progression.