Protein O-GlcNAcylation coupled to Hippo signaling drives vascular dysfunction in diabetic retinopathy

Abstract

Metabolic disorder significantly contributes to diabetic vascular complications, including diabetic retinopathy, the leading cause of blindness in the working-age population. However, the molecular mechanisms by which disturbed metabolic homeostasis causes vascular dysfunction in diabetic retinopathy remain unclear. O-GlcNAcylation modification acts as a nutrient sensor particularly sensitive to ambient glucose. Here, we observe pronounced O-GlcNAc elevation in retina endothelial cells of diabetic retinopathy patients and mouse models. Endothelial-specific depletion or pharmacological inhibition of O-GlcNAc transferase effectively mitigates vascular dysfunction. Mechanistically, we find that Yes-associated protein (YAP) and Transcriptional co-activator with PDZ-binding motif (TAZ), key effectors of the Hippo pathway, are O-GlcNAcylated in diabetic retinopathy. We identify threonine 383 as an O-GlcNAc site on YAP, which inhibits its phosphorylation at serine 397, leading to its stabilization and activation, thereby promoting vascular dysfunction by inducing a pro-angiogenic and glucose metabolic transcriptional program. This work emphasizes the critical role of the O-GlcNAc-Hippo axis in the pathogenesis of diabetic retinopathy and suggests its potential as a therapeutic target.

Fig. 1. Elevated O-GlcNAc-modified protein levels in diabetic retinopathy.

A ELISA analysis of UDP-GlcNAc levels in the vitreous fluid of proliferative diabetic retinopathy (PDR) and non-diabetic patients (n = 7/11 patients). Values are shown as mean ± SEM, with statistical significance at p < 0.0001 (two-tailed Student’s t-test). B Schematic of the streptozotocin (STZ)-induced diabetic model. The upper timeframe represents postnatal weeks, while the lower depicts the duration of diabetes induction by STZ. C Schematic diagram illustrating the oxygen-induced retinopathy (OIR) mouse model. RA: room air. D Western blot and quantification showing elevated O-GlcNAc-modified protein levels in the retina of diabetic mice compared to controls (n = 5/5 mice). Values as mean ± SEM, p = 0.0478 (two-tailed Student’s t-test). E Western blot and quantification showing O-GlcNAc-modified proteins in OIR pups compared to controls (n = 6/6 pups). Values as mean ± SEM, p = 0.046 (two-tailed Student’s t-test). F, G Flat-mounted retinas stained for O-GlcNAc (green) and CD31/IsoB4 (red), illustrating O-GlcNAc levels in control and diabetic/OIR mice. Representative images of O-GlcNAc (green), CD31 (red), and DAPI (blue) staining in frozen eye sections (H) and quantification of O-GlcNAc-modified protein levels in retinal vessels (CD31⁺ area) (I) from diabetic and control mice at different time points (n = 6/5, 6/5, 6/6 mice). Mean ± SEM, p-values < 0.0001 (two-tailed Student’s t-test). Representative images (J) and quantification (K) of O-GlcNAc-modified proteins in retinal vessels (IsoB4⁺ area) from OIR pups and controls (n = 6/6 mice). Mean ± SEM, p = 0.0071 (two-tailed Student’s t-test). L Western blot and quantification of O-GlcNAc-modified proteins in HRCECs exposed to varying glucose concentrations (n = 3 independent experiments). Mean ± SEM, p = 0.0054, p = 0.0007 (one-way ANOVA with Dunnett’s test). M Western blot and quantification of O-GlcNAc-modified proteins in HRCECs under hypoxia for 0, 12, and 24 h (n = 3 independent experiments). Mean ± SEM, p = 0.0051, p = 0.0007 (one-way ANOVA with Dunnett’s test). *p < 0.05; **p < 0.01; ***p < 0.001; scale bars: 100 µm (F, G), 30 µm (H, J). Source data provided as a Source Data file.

Fig. 2. O-GlcNAcylation in ECs affects physiological and pathological retinal angiogenesis.

A BrdU incorporation in HRCECs treated with DMSO or 25 μM PUGNAc for 24 h. B Quantification of BrdU⁺ cells (red) as a percentage of total cell number in (A) (n = 3/3 independent experiments). Mean ± SEM, p = 0.0249 (two-tailed Student’s t-test). C Representative images of mouse aortic rings treated with DMSO or 25 μM PUGNAc for 5 days, stained with Phalloidin (red). D Quantification of microvascular sprouting area / total ring area in (C) (n = 6/7 aortic rings). Mean ± SEM, p = 0.0271 (two-tailed Student’s t-test). E Paracellular resistance in HRCECs treated with DMSO or 25 μM PUGNAc for 12 h. (n = 4/4 independent experiments). Multiple unpaired t-tests using two-stage step-up (Benjamini, Krieger, and Yekutieli) method. F Schematic of STZ and 4-OHT injections in Ogt^WT and Ogt^iΔEC mice. G Desmin (Cyan) and CD31 (Magenta) staining of retinas 9 weeks after STZ onset. Arrowheads indicate vessels without mural cell coverage. H Quantification of mural cell coverage (%) in (G) (n = 6/6 mice). Mean ± SEM, p = 0.0038 (two-tailed Student’s t-test). I Retinal trypsin digestion showing acellular capillaries (Ac-Cap) in STZ mice. Black arrowheads indicate acellular capillaries. J Quantification of Ac-Cap in (I). Ac-Caps were quantified in 3 random fields per retina (n = 6/6 retinas). Mean ± SEM, p = 0.0018 (two-tailed Student’s t-test). K Collagen-lV (green) and CD31 (red) staining of retinas from (G). Arrowheads indicate CollV⁺CD31^- sleeves. L Quantification of CollV⁺CD31^- sleeves per field of retina (n = 3/3 mice, 15 fields per retina). Mean ± SEM, p = 0.0311 (two-tailed Student’s t-test). M Schematic of tamoxifen injections in the OIR model. N Representative images of retinal vasculature stained with IsoB4 (red) in OIR pups at P17. Quantification of retinal neovascularization (O) and avascular area (P) in (N) (n = 13/8 pups). Mean ± SEM, p = 0.0489 (two-tailed Student’s t-test). *p < 0.05; **p < 0.01. Scale bars: 200 µm in (A), (C), 100 µm in (G), 50 µm in (I), (K), and 500 µm in (N). Source data are provided as a Source Data file.

Fig. 3. O-GlcNAcylation regulates Hippo pathway in ECs.

A HRCECs were treated with DMSO or 25 μM PUGNAc for 24 h, and total RNA was harvested for RNA-seq. The enriched pathways for significantly upregulated genes under PUGNAc treatment are shown. The p-value for the Hippo-Yap signaling pathway (WP4537) is 0.0005, computed using the Fisher exact test. B, C HRCECs were infected with Ad-Flag-YAP and Ad-Flag-TAZ for 24 h, followed by treatment with 5.5 mM or 25 mM glucose for an additional 24 h. YAP and TAZ O-GlcNAcylation was analyzed via immunoprecipitation using anti-Flag beads and western blot (n = 3 independent experiments). D, E YAP was purified from HRCECs and analyzed by MS to identify the O-GlcNAcylation sites. Two different O-GlcNAcylation sites in YAP are shown (T358 and T383). F HEK 293T cells were transfected with Flag-YAP-T358A, Flag-YAP-T383A, or Flag-YAP-2TA (T358A and T383A) plasmids. YAP O-GlcNAcylation was analyzed via immunoprecipitation followed by western blot (n = 3 independent experiments). Mean ± SEM, p-values = 0.4728, 0.0488, and 0.0221 (one-way ANOVA with Dunnett’s multiple comparisons test). G Identification of the O-GlcNAcylation modification sites of YAP by in vitro O-GlcNAcylation assay. Purified wild-type or mutant YAP was used as substrates. Coomassie blue staining shows the YAP protein used in the assay. Three independent experiments were performed. *p < 0.05. Source data are provided as a Source Data file.

Fig. 4. YAP O-GlcNAcylation regulates its phosphorylation at S397.

A HEK 293T cells transfected with indicated plasmids. pYAP-S397 was analyzed via immunoprecipitation followed by western blot (n = 3 independent experiments). Mean ± SEM, p-values = 0.0068, 0.004 (one-way ANOVA, Dunnett’s test). B Western blot and quantification of pYAP-S397 in HRCECs treated with DMSO or 25 μM PUGNAc for 24 h (n = 3 independent experiments). Mean ± SEM, p = 0.0002 (two-tailed Student’s t-test). C pYAP-S397 in HRCECs treated with DMSO or 15 μM OSMI-1 for 24 h in 25 mM glucose medium (n = 3 independent experiments). Mean ± SEM, p = 0.0053 (two-tailed Student’s t-test). D, E YAP O-GlcNAcylation in HRCECs infected with Ad-Flag-YAP, transfected with siCtrl or siOGT/siOGA under 25 mM glucose (n = 3 independent experiments). Mean ± SEM, p-values: (D) = 0.0091, 0.0318; (E) = 0.0336, 0.0339 (two-tailed Student’s t-test). F, G Relative CTGF and CYR61 mRNA in HRCECs treated as in (D) analyzed by qRT-PCR (n = 3 independent experiments). Mean ± SEM, p-values: (F) = 0.0191, 0.0458; (G) = 0.0298, 0.0443 (two-tailed Student’s t-test). H HRCECs infected with indicated adenovirus, treated with DMSO or 25 μM PUGNAc, and analyzed for pYAP-S397 (n = 3 independent experiments). Mean ± SEM, p = 0.0012 (one-way ANOVA, Dunnett’s test). I Ubiquitination and β-TRCP levels in HRCECs infected with indicated adenovirus and treated with MG132 or vehicle (n = 3 independent experiments). Mean ± SEM, p-values = 0.0011, 0.0002, 0.0196, 0.0125 (one-way ANOVA, Tukey’s test). J YAP stability in HRCECs infected with indicated adenovirus and treated with 50 µg/ml CHX for western blot. Flag-YAP protein levels were quantified as Flag-YAP/β-Actin (n = 3 independent experiments). Mean ± SEM, p-values: 0.0064, <0.0001 (two-way ANOVA, Šídák’s test). K–N Images and quantification of YAP/TAZ (green) within vessels (CD31⁺ or IsoB4⁺ areas, red) from Ogt^WT and Ogt^iΔEC STZ or OIR mice (n = 4/4 mice in L, 6/6 mice in N). Mean ± SEM, p-values: (L) = 0.0443; (N) = 0.0002 (two-tailed Student’s t-test). *p < 0.05; **p < 0.01; ***p < 0.001. Scale bars: 30 µm in (K, M). Source data provided as a Source Data file.

Fig. 5. Endothelial YAP/TAZ expression is elevated in DR and plays a critical role in pathological angiogenesis.

A Proximity ligation assay (PLA) of YAP/TAZ and OGT interaction in frozen eyeball sections from control and STZ-induced diabetic mice. CD31 was stained for blood vessels. B Quantification of PLA counts (red) in CD31⁺ areas (green) (n = 3/3 mice). Mean ± SEM, p = 0.016 (two-tailed Student’s t-test). C PLA of YAP/TAZ and OGT interaction in frozen eyeball sections from control and OIR model pups. CD31 was stained for blood vessels. D Quantification of PLA counts in CD31⁺ areas (n = 3/3 mice). Mean ± SEM, p = 0.0033 (two-tailed Student’s t-test). E Immunostaining of YAP/TAZ (green), UEA1 (red), and DAPI (blue) in paraffin-embedded eyeball sections from non-diabetic control and PDR patients. Arrowheads indicate YAP/TAZ localization in EC nuclei. F Quantification of YAP/TAZ fluorescence intensity within vessels (UEA1⁺ areas) from non-diabetic controls and PDR patients (n = 4/4 individuals). Mean ± SEM, p = 0.008 (two-tailed Student’s t-test). G Relative expression of CTGF and CYR61 in retinal microvascular ECs from PDR patients compared to non-diabetic controls, analyzed from the Omnibus Database (GSE94019) (n = 4/9 individuals). Mean ± SEM, p = 0.0268 (two-tailed Student’s t-test). H Immunostaining of YAP/TAZ (green) and CD31 (red) in sagittal retinal sections from STZ-induced diabetic mice at different stages and age-matched controls. I Quantification of YAP/TAZ fluorescence intensity in CD31⁺ areas (n = 3/3 mice per time point). Mean ± SEM, p-values = 0.0175, 0.0008, 0.0007 (two-tailed Student’s t-test). J Immunostaining of YAP/TAZ (green) and IsoB4 (red) in sagittal retinal sections from OIR model mice and age-matched controls. K Quantification of YAP/TAZ fluorescence intensity in IsoB4⁺ areas (n = 6/6 mice). Mean ± SEM, p < 0.0001 (two-tailed Student’s t-test). *p < 0.05; **p < 0.01. Scale bars: 5 µm in (A, C); 30 µm in (E, H, J). Source data are provided as a Source Data file.

Fig. 6. Endothelial YAP/TAZ is essential for regulating pathological retinal angiogenesis.

A Schematic showing the generation of EC-specific Yap/Taz knockout mice. B Intravitreal 4-OHT injection and STZ administration in EC-specific Yap/Taz knockout (Yap/Taz^iΔEC) and littermate control (Yap/Taz^WT) mice. C Desmin (Cyan) and CD31 (magenta) staining in retinas of Yap/Taz^WT and Yap/Taz^iΔEC mice, 9 weeks post-STZ onset. Arrowheads indicate vessels lacking mural cell coverage. D Quantification of mural cell coverage (%) within CD31⁺ vessel areas from (C) (n = 6/6 mice). Mean ± SEM, p = 0.0484 (two-tailed Student’s t-test). E Retinal trypsin digestion showing acellular capillaries in Yap/Taz^WT and Yap/Taz^iΔEC STZ mice. Black arrowheads indicating acellular capillaries. F Acellular capillaries (Ac-Cap) in (E), based on 3 random fields per retina (n = 6/6 retinas). Mean ± SEM, p = 0.0484 (two-tailed Student’s t-test). G Collagen lV (green) and CD31 (red) staining in retinas, 9 weeks post-STZ-induced diabetes. Arrowheads indicate Ac-Caps. H Quantification of CollV⁺CD31^- sleeves in (G) (15 random fields per retina, n = 3/3 mice). Mean ± SEM, p = 0.0479 (two-tailed Student’s t-test). I Schematic of Tamoxifen injection in the OIR model. J IsoB4 staining of retinal vasculature in P17 OIR retinas from Yap/Taz^WT and Yap/Taz^iΔEC pups. K, L Quantification of neovascularization and avascular areas in Yap/Taz^WT and Yap/Taz^iΔEC OIR retinas (n = 7/8 pups). Mean ± SEM, p < 0.0001 in (K), p = 0.9361 in (L) (two-tailed Student’s t-test). M, N Images and quantification of blood island areas in P17 OIR retinas (n = 10/14 pups). Mean ± SEM, p = 0.0013 (two-tailed Student’s t-test). O, P Images and quantification of TER119⁺ red blood cell (RBC) (Cyan) leakage in flat-mounted retinas of OIR pups (n = 8/9 pups). Mean ± SEM, p = 0.0003 (two-tailed Student’s t-test). Q, R Images and quantification of extravasated FITC-dextran (70 kDa) (green) in IsoB4-stained (red) flat-mounted retinas (n = 5/8 pups). Mean ± SEM, p = 0.001 (two-tailed Student’s t-test). *p < 0.05; **p < 0.01; ***p < 0.001. Scale bars: 100 µm in (C), 50 µm in (G), 500 µm in (J, O, Q), and 1 mm in (M). Source data are provided as a Source Data file.

Fig. 7. YAP-T383/YAP-S397 regulatory axis is essential in O-GlcNAcylation-modulated pathological retinal angiogenesis.

A Schematic showing AAV-BR1 injection in the OIR model. AAV-BR1 was injected retro-orbitally at P7, followed by tamoxifen intraperitoneal injections at P12, P13, P14, and P16 in Yap/Taz^iΔEC pups. B IsoB4-stained images of retinal vasculature in OIR retinas from AAV-BR1-injected Yap/Taz^iΔEC pups at P17. C Quantification of retinal neovascularization area in (B) (n = 6/6/7 pups). Values are mean ± SEM, p = 0.0023, p = 0.0097 (one-way ANOVA with Tukey’s multiple comparisons test). D Schematic of AAV-BR1 injection in the STZ model. One week after STZ onset, 4-OHT was injected intravitreally into Yap/Taz^iΔEC mice, with AAV-BR1 injected into the tail vein. Retinas were harvested after 8 weeks for analysis. E Desmin (Cyan) and CD31 (magenta) staining of retinas from AAV-BR1-injected Yap/Taz^iΔEC mice. Arrowheads show vessels lacking mural cell coverage. F Quantification of mural cell coverage (%) in CD31⁺ vessel areas from (E) (n = 6/6/5 mice). Values are mean ± SEM, p = 0.003, p = 0.0424 (one-way ANOVA with Tukey’s test). G Schematic of AAV-BR1 injection in STZ model. 4-OHT was intravitreally injected into Ogt^WT and Ogt^iΔEC mice one week after STZ onset, with AAV-BR1 injected intravenously. Retinas were harvested after 8 weeks. H Desmin (Cyan) and CD31 (magenta) staining of retinas from AAV-BR1-injected Ogt^WT and Ogt^iΔEC mice. Arrowheads indicate vessels lacking mural cell coverage. I Quantification of mural cell coverage (%) in CD31⁺ vessel areas from (H) (n = 8/7/7/9/8 mice). Values are mean ± SEM, p = 0.0051, p = 0.0002, p < 0.0001 (one-way ANOVA with Tukey’s test). J Schematic of AAV-BR1 injection in OIR model, with AAV-BR1 injected retro-orbitally at P7 into Ogt^WT and Ogt^iΔEC pups. K Representative IsoB4-stained images of retinal vasculature from AAV-BR1-injected Ogt^WT and Ogt^iΔEC OIR pups. L Quantification of retinal neovascularization in AAV-BR1-injected Ogt^WT and Ogt^iΔEC pups (n = 5/6/6/6/5 pups). Values are mean ± SEM, p = 0.0059, p = 0.0018, p < 0.0001 (one-way ANOVA with Tukey’s test). *p < 0.05; **p < 0.01; ***p < 0.001. Scale bars: 500 µm in (B, K), 100 µm in (E, H). Source data are provided as a Source Data file.

Fig. 8. Exosome-based delivery of YAP/TAZ inhibitor alleviates vascular dysfunction in DR.

A Schematic showing the preparation of EXO_Super-TDU, where Super-TDU is anchored to EC-derived exosomes using the CP05 peptide. B Schematic of Super-TDU and EXO_Super-TDU administration in the STZ model. PBS, Super-TDU, or EXO_Super-TDU were injected retro-orbitally one week after STZ onset, with vascular phenotypes analyzed 8 weeks later. C Desmin (Cyan) and CD31 (magenta) staining of retinas from STZ mice treated as indicated. Arrowheads indicate vessels lacking mural cell coverage. D Quantification of mural cell coverage (%) in CD31⁺ vessels from (C) (n = 6 mice per group). Mean ± SEM, p = 0.0319, p < 0.0001, p = 0.0217 (one-way ANOVA with Tukey’s test). E Retinal trypsin digestion showing acellular capillaries in STZ mice. Black arrowheads indicate acellular capillaries. F Quantification of acellular capillaries (Ac-Cap) in (E). Three random fields per retina were quantified (n = 6 mice per group). Mean ± SEM, p = 0.0108, p < 0.0001, p = 0.0177 (one-way ANOVA with Tukey’s test). G Schematic of Super-TDU treatment in the OIR model. PBS, Super-TDU, or EXO_Super-TDU were injected retro-orbitally at P12, and retinas were harvested at P17. H IsoB4-stained images of retinal vasculature in OIR retinas from treated pups. I, J Quantification of neovascularization and avascular areas in OIR retinas (n = 6/6 pups). Mean ± SEM, p = 0.0349, p = 0.0002, p = 0.0499 (one-way ANOVA with Tukey’s test). K, L Images of the retinal cup inner surface and quantification of blood island area (n = 8/7/7 pups). Mean ± SEM, p = 0.003, p < 0.0001, p = 0.0374 (one-way ANOVA with Tukey’s test). M, N Confocal images and quantification of extravasated FITC-dextran and IsoB4 in flat-mounted OIR retinas. FITC-dextran⁺ and IsoB4⁻ areas indicate vessel leakage (n = 6 pups per group). Mean ± SEM, p = 0.0487, p < 0.0001, p = 0.0074 (one-way ANOVA with Tukey’s test). *p < 0.05; **p < 0.01; ***p < 0.001. Scale bars: 100 µm (C), 50 µm (E), 1 mm (K), 500 µm (H, M). Source data are provided as a Source Data file.

Fig. 9. YAP/TAZ regulates glucose metabolism and O-GlcNAcylation in ECs.

A HRCECs were transfected with siCtrl or siYAP/TAZ and cultured in 25 mM glucose medium for 48 h. Total RNA was then harvested for RNA-seq. B Network visualization of GO terms for genes downregulated by YAP/TAZ knockdown. C Schematic representation of the hexosamine biosynthetic pathway (HBP), with key enzymes and O-GlcNAcylation processes highlighted in yellow boxes. D Heatmap of RNA-seq data from (A), displaying expression levels of key enzymes involved in HBP and O-GlcNAcylation. E Relative expression levels of key enzymes in the HBP and O-GlcNAcylation in retinal microvascular ECs from PDR patients compared to non-diabetic controls. Data from GSE94019 (n = 4/9 individuals). Mean ± SEM, p = 0.0466, p = 0.0234 (multiple unpaired Student’s t-tests). F Western blot and quantification of overall O-GlcNAc-modified proteins in HRCECs transfected with siCtrl or siYAP/TAZ, followed by infection with Ad-Flag-YAP/TAZ and cultured in 25 mM glucose medium (n = 3 independent experiments). Mean ± SEM, p = 0.036, p = 0.0341 (one-way ANOVA with Tukey’s test). G, H O-GlcNAc (green) and CD31 (red) staining with DAPI (blue) in frozen sections of eyeballs from Yap/Taz^WT and Yap/Taz^iΔEC mice 9 weeks after the onset of STZ-induced diabetes (n = 8/6 mice). Mean ± SEM, p = 0.002 (two-tailed Student’s t-test). I, J O-GlcNAc and IsoB4 (red) staining with DAPI (blue) in frozen sections of eyeballs from Yap/Taz^WT and Yap/Taz^iΔEC OIR pups at P17 (n = 4/5 pups). Mean ± SEM, p = 0.0084 (two-tailed Student’s t-test). K The proposed working model demonstrates the role of O-GlcNAcylation in regulating vascular dysfunction in DR through the modulation of YAP/TAZ. Exposure to high glucose or hypoxia results in O-GlcNAcylation of YAP/TAZ at site T383, which impedes phosphorylation at S397, leading to the stabilization and activation of YAP/TAZ. This modification ultimately contributes to vascular dysfunction in DR. Additionally, YAP/TAZ also exerts regulatory control over EC metabolism and protein O-GlcNAcylation, establishing a reciprocal interplay between the Hippo signaling pathway and protein O-GlcNAcylation. *p < 0.05; **p < 0.01. Scale bars: 30 µm in (G) and (I).