Essential Role of Transglutaminase 2 in Vascular Endothelial Growth Factor–Induced Vascular Leakage in the Retina of Diabetic Mice

Hyperglycemia and subsequent metabolic changes develop into diabetic complications as a result of progressive damage to and the dysfunction of blood vessels (1). Diabetic retinopathy is a major microvascular complication and is the leading cause of blindness in adults (1–3). Diabetic retinopathy has two pathogenic processes: nonproliferative diabetic retinopathy in the early stages and proliferative diabetic retinopathy in the later stages (4–6). In nonproliferative diabetic retinopathy, retinal blood vessels are damaged by sustained hyperglycemia, resulting in blood flow alterations, retinal pericyte loss, microaneurysm, basement membrane thickening, and vascular leakage (7,8). In proliferative diabetic retinopathy, a number of the blood vessels are blocked, neovascularization starts as a result of hypoxia, and severe vision loss occurs via retinal detachment (4,7,9). Thus prevention of vasculature alteration and leakage in the early stages is necessary to prevent diabetic retinopathy.

The retinal vascular leakage in the early stages of diabetic retinopathy is predominantly caused by vascular endothelial growth factor (VEGF)–mediated stress fiber formation and vascular endothelial (VE)–cadherin disruption (10,11). Retinopathy develops in nearly all patients with type 1 diabetes and in >60% of those with type 2 diabetes (1,6). Upregulation of VEGF and its receptors by hyperglycemic conditions elevates intracellular reactive oxygen species (ROS) levels (7,11,12). Intracellular ROS generated directly or indirectly lead to the loss of stability and the internalization of VE-cadherin through microfilament reorganization (11). However, the molecular mechanism by which VEGF disrupts vascular integrity has not been clearly elucidated.

Transglutaminase 2 (TGase2), also known as tissue transglutaminase, is a member of the TGase family and catalyzes protein cross-linking reactions through the transamidation of glutamine residues to lysine residues in a Ca²⁺-dependent manner (13). TGase2 is expressed ubiquitously and is implicated in various cellular processes including cell death, proliferation, adhesion, migration, and cytoskeletal reorganization (13–15). TGase2 is also associated with various diseases, including cardiovascular diseases (16), inflammation (17,18), celiac disease (19,20), neurodegenerative disorders (21,22), and hemorrhagic or ischemic stroke (23). TGase2 is expressed in ocular tissues, including the retina and lens, and is possibly associated with ocular diseases including cataract and glaucoma (18,24,25). However, the role of TGase2 in the vascular leakage involved in the pathogenesis of microvascular diabetic complications, including retinopathy, is not clearly understood.

In this study we aimed to elucidate the molecular mechanism by which VEGF disrupts adherens junctions and induces vascular leakage in the retina of diabetic mice. We hypothesized that ROS-mediated TGase2 activation plays an essential role in VEGF-induced vascular leakage in the diabetic retina. We designed an in vivo TGase activity assay in mouse retina and demonstrated that hyperglycemia induced vascular leakage by activating TGase2 in diabetic retina. TGase inhibitors or TGase2 small interfering RNA (siRNA) prevented VEGF-induced stress fiber formation and VE-cadherin disruption in human retinal endothelial cells (HRECs). Our in vitro findings were validated using streptozotocin-induced diabetic mice. We found that in vivo TGase activity was elevated in the diabetic retina, and intravitreal injections of TGase inhibitors and TGase2 siRNA prevented hyperglycemia-induced TGase activation and vascular leakage in the diabetic retina. The role of TGase2 in VEGF-induced vascular leakage was further demonstrated in diabetic TGase2^−/− mice. Our findings suggest that TGase2 plays a pivotal role in VEGF-induced vascular integrity disruption and vascular leakage in the diabetic retina.

HRECs were purchased from the Applied Cell Biology Research Institute (Cell Systems, Kirkland, WA) and grown on 2% gelatin-coated plates in M199 medium supplemented with 20% FBS, 3 ng/mL basic fibroblast growth factor, 5 U/mL heparin, 100 U/mL penicillin, and 100 μg/mL streptomycin. For experiments, cells were incubated for 6 h in low-serum medium supplemented with 1% FBS and antibiotics.

Changes in intracellular Ca²⁺ levels were monitored as previously described (11). Cells on coverslips were incubated with 2 μmol/L Fluo4-AM and various inhibitors at 37°C for 30 min. Coverslips were scanned every 10 s in the presence of 10 ng/mL VEGF by a confocal microscope (FluoView 300; Olympus, Tokyo, Japan). Peaks of single-cell fluorescence intensities were determined for 10 cells/experiment (selected randomly).

Intracellular ROS generation was determined using 2′,7′-dichlorodihydrofluorescein diacetate (Molecular Probes, Eugene, OR) staining, as previously described (26). Cells were pretreated with the inhibitors for 30 min and incubated with 10 ng/mL VEGF and 10 μmol/L dichlorodihydrofluorescein diacetate in low-serum media (phenol red-free) for 10 min. Labeled cells were immediately analyzed by confocal microscopy (FluoView 300; Olympus). Single-cell fluorescence intensities were determined for 10 cells/experiment (selected randomly). The levels of intracellular ROS were determined by comparing the average fluorescence intensities of treated cells with those of control cells (fold change).

In situ TGase transamidating activity was determined by confocal microscopic assay according to a previously published procedure (27). Briefly, cells were incubated with 1 mmol/L 5-(biotinamido)pentylamine (BAPA) at 37°C for the last 1 h, fixed with 3.7% formaldehyde in PBS for 30 min, and permeabilized with 0.2% Triton X-100 in PBS for 30 min. After incubation with a blocking solution of 2% BSA in 20 mmol/L Tris (pH 7.6), 138 mmol/L NaCl, and 0.1% Tween 20 for 30 min, cells were treated with Cy3-conjugated streptavidin (1:200, v/v) in the blocking solution for 1 h. Fluorescence intensities of single stained cells were determined by confocal microscopy 10 cells/experiment, randomly selected. TGase activity was determined by comparing the average fluorescence intensities of treated cells with those of control cells (fold change).

HRECs were transfected with human TGase2-specific siRNA as described previously (28). Briefly, cells were transfected with various concentrations of human TGase2-siRNA or 100 nmol/L control siRNA (Dharmacon, Inc., Lafayette, CO) using siLentFect lipid reagent (Bio-Rad Laboratories, Hercules, CA) according to the manufacturer’s instructions. Six hours after transfection with the indicated concentrations, the media was replaced with fresh culture medium and the transfected cells were incubated for another 24 h.

Cells were incubated with ice-cold lysis buffer containing 50 mmol/L Tris-HCl (pH 7.5), 1% Triton X-100, 150 mmol/L NaCl, 1 mmol/L EDTA, 0.1 mmol/L phenylmethylsulfonyl fluoride, 10 μg/mL aprotinin, and 10 μg/mL leupeptin for 30 min. Cell lysates were separated by SDS-PAGE and transferred to polyvinylidene fluoride membranes. The membranes were blocked with 5% skim milk in Tris-buffered saline with 0.1% Tween for 1 h at room temperature. The membranes were probed with TGase2 antibody (Santa Cruz Biotechnology, Santa Cruz, CA) and phospho-Src (Tyr⁴¹⁶) antibody (Cell Signaling Technology, Danvers, MA), followed by incubation with a horseradish peroxidase–conjugated secondary antibody. Protein bands were visualized using a chemiluminescent substrate (AbFrontier, Seoul, Korea).

Actin filaments and VE-cadherin were visualized as previously described (11). For visualization of actin filaments, cells were treated with 10 ng/mL VEGF at 37°C for 1 h, fixed with 3.7% formaldehyde, and permeabilized with 0.2% Triton X-100. Cells were incubated with Alexa Fluor 488 phalloidin (1:200; Molecular Probes) for 1 h at room temperature, and actin filaments were observed with a confocal microscope. For visualization of VE-cadherin, cells were incubated with the TGase2 inhibitors for 30 min and treated with 10 ng/mL VEGF at 37°C for 90 min. Following fixation and permeabilization, cells were incubated overnight with a monoclonal VE-cadherin antibody (1:200; Santa Cruz Biotechnology) at 4°C. Cells were then probed with a fluorescein isothiocyanate (FITC)–conjugated goat antimouse antibody (1:200; Sigma-Aldrich), and VE-cadherin was visualized using confocal microscopy. Adherens junctions are represented by histograms of VE-cadherin (as indicated by dotted lines) (Figs. 2G and 3B) and were quantitatively analyzed using peak fluorescence intensities of the histograms at the single-cell level for 10 cells/experiment (randomly selected).

For visualization of β-catenin, cells were incubated with inhibitors for 30 min and treated with 10 ng/mL VEGF for 90 min, fixed with 3.7% formaldehyde, and permeabilized with 0.2% Triton X-100. Cells were incubated overnight with a monoclonal β-catenin antibody (1:200; Santa Cruz Biotechnology) at 4°C. Cells were then probed with a FITC-conjugated goat antimouse antibody (1:200; Sigma-Aldrich) and visualized using confocal microscopy. Images were cropped into smaller size using PowerPoint 2010 software (Microsoft Corp., Redmond, WA) to make image galleries.

In vitro permeability assay was performed as previously described (11). HRECs were grown on gelatin-coated inserts (0.4-μm polycarbonate membrane) of Transwell Permeable Supports (Costar, Corning, NY) up to confluence. Approximately 1.0 × 10⁵ cells in 0.5 mL of culture medium were seeded at the upper side of the membrane, whereas 1 mL of culture medium was added to the lower compartment. After culturing for 5 days, cells on the inserts were incubated with 50 μmol/L cystamine, 20 μmol/L monodansylcadaverine (MDC), or 100 nmol/L dasatinib (BioVision, Milpitas, CA) for 30 min, treated with 10 ng/mL VEGF for 90 min, and incubated with 1 mg/mL 40-kDa FITC-dextran (Sigma-Aldrich) for the last 60 min. The amount of FITC-dextran diffused through the endothelial monolayer into the lower chamber was measured by a microplate spectrofluorometer (Molecular Devices, Sunnyvale, CA).

Six-week-old male C57BL/6 mice were obtained from KOATECH (Pyeongtaek, Korea). TGase2^−/− mice (C57BL/6) were prepared by disrupting exons 5 and 6 of TGase2 by homologous recombination, as previously described (29). Diabetic mice were generated by a single intraperitoneal injection of streptozotocin (150 mg/kg body weight; Sigma-Aldrich) freshly prepared in 100 mmol/L citrate buffer (pH 4.5), as previously described (30). Genotyping and clinical data of TGase2^−/− mice are shown in Supplementary Fig. 1. Mice with nonfasting blood glucose concentrations ≥19 mmol/L, polyuria, and glucosuria were considered diabetic. Experiments were performed in strict accordance with the guidelines of the Institutional Animal Care and Use Ethics Committee of Kangwon National University.

Two weeks after streptozotocin injection, diabetic mice were intravitreally injected in one eye with 2 μL TGase inhibitors (50 mmol/L cystamine or 20 mmol/L MDC); 1 μmol/L C-peptide (American Peptide Co., Sunnyvale, CA); ROS scavengers (2 μmol/L Trolox or 500 mmol/L N-acetyl cysteine [NAC]); or 65 μmol/L mouse TG2-specific siRNA (Dharmacon, Inc.), and an equal volume of PBS or mouse control siRNA (Dharmacon, Inc.) was injected into the contralateral eye. Nondiabetic mice were also intravitreally injected with 2 μL PBS or mouse control siRNA into the eyes. After 24 h for inhibitors or C-peptide or 48 h for mouse TGase2 siRNA, in vivo TGase activity and vascular leakage in the mouse retinas were measured.

In vivo TGase transamidating activity in mouse retina was determined by confocal microscopy (Fig. 5A). Twenty-four hours after intravitreal injection of TGase inhibitors and C-peptide or 48 h after injection of mouse TGase2-specific siRNA, mice were deeply anesthetized using 2.5% avertin, and 48 μL of 100 mmol/L BAPA was injected into the left ventricle. BAPA was allowed to circulate for 10 min, then mice were killed by cervical dislocation. Eyes were enucleated and immediately fixed with 4% paraformaldehyde for 45 min. Retinas were dissected in the Maltese cross-configuration and permeabilized with 0.2% Triton X-100 in PBS for 30 min at room temperature. After 30 min of incubation with the blocking solution, retinas were treated with FITC-conjugated streptavidin (1:200, v/v) in the blocking solution for 1 h at room temperature. The superficial vessels of stained retinas (n = 8 per group) were observed using confocal microscopy, and fluorescence intensities were analyzed by the FluoView software (FluoView 300). In vivo TGase activities were quantitatively determined using fluorescence intensities in the retinas of normal and diabetic mice.

Microvascular leakage in the mouse retinas was investigated using fluorescein angiography, as previously described (11). Briefly, 24 or 48 h after intravitreal injection of TGase inhibitors or mouse TGase2 siRNA, respectively, mice were deeply anesthetized, and 1.25 mg of 500-kDa FITC-dextran (Sigma-Aldrich) was injected into the left ventricle. The dye was allowed to circulate for 5 min, then mice were killed by cervical dislocation. The eyes were enucleated and immediately fixed with 4% paraformaldehyde for 45 min. Retinas were dissected in the Maltese cross-configuration and flat-mounted onto glass slides. The superficial vessels of the retinas (n = 8 per group) were observed by confocal microscopy, and vascular leakage was quantitatively analyzed using the FluoView software by determining fluorescence intensities of FITC-dextran extravasated from the retina vessel.

For staining actin filaments, mice were killed by cervical dislocation and eyes were enucleated and immediately fixed overnight with 4% paraformaldehyde at 4°C. Following dissection in the Maltese cross-configuration, retinas were permeabilized with 1.0% Triton X-100 in PBS for 1 h and incubated with Alexa Fluor 488 phalloidin (1:200; Molecular Probes) for 2 h at room temperature. For staining VE-cadherin, enucleated eyes were fixed with 4% paraformaldehyde for 45 min at room temperature and acetone for 3 min at −20°C. Following dissection in the Maltese cross-configuration, retinas were permeabilized with 1.0% Triton X-100 in PBS for 4 h at room temperature and incubated overnight with a monoclonal VE-cadherin antibody (1:100; BD Pharmingen, San Diego, CA) at 4°C. Retinas were then incubated with a FITC-conjugated goat antimouse antibody (1:300; Sigma-Aldrich) for 2 h at room temperature. Actin filaments and VE-cadherin in the superficial vessels of retinas were visualized by confocal microscopy.

Mouse eyes were fixed overnight with 4% paraformaldehyde at 4°C, incubated overnight with 30% sucrose at 4°C, and frozen in optimal cutting temperature compound. Frozen sections (10 μm) were cut using a microtome-cryostat (Leica Biosystems, Wetzlar, Germany) and washed three times with PBS. Antigen was retrieved with 0.01 mol/L citrate buffer (pH 6.0) for 20 min at 95°C. Following cooling for 20 min at room temperature, sections were rinsed three times with 0.1% Tween 20 in PBS and incubated with the blocking solution for 30 min. Sections were then incubated overnight with TGase2 antibody (1:200; Santa Cruz Biotechnology) in the blocking solution and probed with a FITC-conjugated goat antirabbit antibody (1:300; Sigma-Aldrich) for 2 h at room temperature. TGase2 expression was visualized using confocal microscopy.

Data processing was performed using Origin 6.1 software (OriginLab, Northampton, MA) and are expressed as the mean ± SD of at least three independent experiments. Statistical significance was determined using ANOVA. P < 0.05 was considered statistically significant.

We performed in situ TGase activity assays using confocal microscopy to study whether VEGF can activate TGase in HRECs. It was reported that vascular leakage in diabetic retina is predominantly caused by VEGF (10,11). VEGF increased in situ TGase activity in an incubation time–dependent manner, with maximal activation at 2 h (P < 0.001) (Fig. 1A). VEGF-induced TGase activation was inhibited by the ROS scavengers NAC and Trolox or the Ca²⁺ chelator BAPTA-AM (P < 0.001) (Fig. 1B), indicating that VEGF can activate TGase via an increase in intracellular ROS and Ca²⁺ levels. As expected, the TGase inhibitors cystamine and MDC prevented VEGF-induced elevation of TGase activity (P < 0.001) (Fig. 1B).

We investigated the role of TGase2 in VEGF-induced TGase activation by transfecting the endothelial cells with human TGase2 siRNA. Human TGase2 siRNA suppressed TGase2 protein expression in a dose-dependent manner, with a maximal effect at 100 nmol/L (Fig. 1C). TGase2 siRNA totally prevented VEGF-induced TGase activation (P < 0.001), whereas the control siRNA had no effect (Fig. 1D). TGase2 siRNA or control siRNA alone showed negligible effects on TGase activity. These results demonstrate that TGase2—and not other members of the TGase family—mostly contributed to a VEGF-induced increase of in situ TGase activity.

We studied the roles of intracellular ROS and Ca²⁺ in VEGF-induced TGase2 activation using various inhibitors in endothelial cells. VEGF increased intracellular Ca²⁺ levels; this increase was inhibited by BAPTA-AM (P < 0.001) but not by the ROS scavengers or the TGase inhibitors (Fig. 1E). VEGF induced intracellular ROS generation, which was significantly inhibited by the ROS scavengers or BAPTA-AM (P < 0.001) but not by the TGase inhibitors (Fig. 1F), suggesting that intracellular Ca²⁺ is involved in VEGF-induced ROS generation. These results elucidate that VEGF activated TGase2 via sequential elevation of intracellular Ca²⁺ and ROS levels in endothelial cells.

To investigate the role of TGase2 in VEGF-mediated vascular leakage, we studied the effects of cystamine and MDC on VEGF-induced stress fiber formation and adherens junction disruption in HRECs. VEGF activated stress fiber formation, which was prevented by the two TGase inhibitors (Fig. 2A). VEGF induced VE-cadherin disassembly, which was inhibited by the TGase inhibitors (Fig. 2B). The changes in the stability of VE-cadherin are represented by line profiles displaying the distribution of relative fluorescence intensity, as shown by dotted lines crossing two cell-cell contacts. VE-cadherin disassembly was also quantitatively analyzed by measuring the peak fluorescence intensities of the histograms (Fig. 2C). The role of TGase in vascular leakage was further investigated by in vitro endothelial cell monolayer permeability assay using HRECs. VEGF increased in vitro endothelial permeability, which was inhibited by the TGase inhibitors (P < 0.01) (Fig. 2D). Thus it is likely that TGase2 is implicated in VEGF-induced endothelial cell permeability through stress fiber formation and VE-cadherin disruption.

To confirm the role of TGase2 in VEGF-induced vascular leakage, we investigated the effect of human TGase2 siRNA on VEGF-induced stress fiber formation and VE-cadherin disruption in endothelial cells. Consistent with the findings in Fig. 2, VEGF activated stress fiber formation in endothelial cells transfected with control siRNA, but not in TGase2 siRNA–treated cells (Fig. 3A). TGase2 siRNA or control siRNA alone had no significant effect on stress fiber formation. VEGF induced VE-cadherin disruption in control siRNA–treated endothelial cells; this disruption was prevented by TGase2 siRNA transfection (P < 0.001) (Fig. 3B and C). TGase2 siRNA or control siRNA alone had no effect on VE-cadherin disruption. Taken together, these data show that TGase2 plays a critical role in VEGF-mediated endothelial permeability through stress fiber formation and VE-cadherin disruption in endothelial cells.

We then investigated whether β-catenin and c-Src are involved in TGase2-mediated endothelial permeability in HRECs. It was reported that TGase regulates β-catenin signaling through a c-Src-dependent mechanism (31). VEGF induced β-catenin disassembly, which was inhibited by the TGase inhibitor cystamine and the Src family kinase inhibitor dasatinib (Fig. 4A). Cystamine prevented c-Src phosphorylation (Tyr⁴¹⁶) induced by VEGF. Dasatinib blocked VRGF-induced c-Src phosphorylation (Fig. 4B) and endothelial permeability (Fig. 4C). These results suggest that TGase2 is involved in VEGF-induced adherens junction disassembly through a mechanism involving C-Src and β-catenin.

To validate our in vitro findings, we further investigated the role of TGase2 in hyperglycemia-induced vascular leakage in the retinas of diabetic mice. For this study, we designed an in vivo TGase activity assay using confocal microscopy of mouse retina (Fig. 5A). In this assay, BAPA, a TGase pseudosubstrate, was systemically delivered into the blood circulation by injection into the left ventricles of mice, and biotinylated proteins in the retinas were probed with FITC-conjugated streptavidin.

In vivo TGase activity was highly elevated in the blood vessels and ganglion cells of streptozotocin-induced diabetic mouse retina compared with the normal retina (n = 8); this hyperglycemia-induced TGase activation was suppressed by intravitreal injection of the TGase inhibitors cystamine and MDC (Fig. 5B and Supplementary Fig. 2). We then quantitatively analyzed in vivo TGase activity by determining the fluorescence intensities of FITC-conjugated streptavidin in retina tissues. The average TGase activity in the retinas of diabetic mice was approximately twofold higher than that of normal mice (P < 0.001; n = 8) (Fig. 5C). The TGase inhibitors reversed the hyperglycemia-induced increase of TGase activity in the blood vessels and ganglion cells of diabetic retinas (P < 0.001; n = 8) (Fig. 5C). These results demonstrate that our in vivo confocal microscopic assay successfully analyzed TGase activation by hyperglycemia in diabetic retina.

We then investigated the role of TGase in vascular leakage in the diabetic retina using fluorescence angiography after intravitreal injection of cystamine and MDC into mouse eyes. High levels of extravasation of FITC-dextran were observed in the retinas of diabetic mice; this leakage was blocked in the retinas of the contralateral eyes injected with the TGase inhibitors (Fig. 6A and Supplementary Fig. 2). Fluorescence intensity in the retinas of diabetic mice was approximately threefold higher than that in control mice (P < 0.001; n = 8); however, cystamine and MDC prevented vascular leakage (P < 0.001; n = 8) (Fig. 6B). These results suggest that hyperglycemia induces vascular leakage by activating TGase in the retinas of diabetic mice.

We investigated whether hyperglycemia can induce stress fiber formation and adherens junction disassembly in diabetic retina. Consistent with in vitro findings, actin filament staining showed the enhanced formation of stress fibers in diabetic retinas compared with normal retinas (Fig. 6C). The VE-cadherin staining showed continuous VE-cadherin distribution along the junctions in the normal retinas, but exhibited frequent disrupted adherens junctions in diabetic retinas (Fig. 6D). These results indicate that hyperglycemia induces vascular leakage by stress fiber formation and adherens junction disruption in the retinas of diabetic mice.

We examined the effect of the ROS scavengers Trolox and NAC on hyperglycemia-induced TGase activation in the diabetic retina. Trolox prevented this (P < 0.001) (Fig. 5C), which is consistent with the in vitro results shown in Fig. 1B. A similar inhibitory effect was observed by NAC (P < 0.01) (Fig. 5C). The role of intracellular ROS in the hyperglycemia-induced TGase activation was further studied using C-peptide because C-peptide prevents vascular permeability by inhibiting VEGF-stimulated ROS generation in the diabetic retina (11). Human C-peptide is a 31–amino acid peptide that is released from β-cells into the peripheral circulation in equimolar concentrations with insulin (1). Intravitreal injection of the C-peptide blocked TGase activation in the diabetic retina (P < 0.001) (Fig. 5B and C). Considering the role of intracellular ROS in VEGF-mediated vascular leakage in the diabetic retina, these results suggest that TGase is activated by VEGF-increased intracellular ROS in the diabetic retina.

To confirm the role of TGase2 in vascular permeability in the diabetic retina, diabetic mice were intravitreally injected with mouse TGase2-specific siRNA, and in vivo TGase activity and vascular leakage in retinas were analyzed. In vivo TGase activity was significantly higher in the diabetic retinas injected with control siRNA compared with the normal retinas (P < 0.001; n = 8); this activity was blocked in the retinas of the contralateral eyes injected with TGase2 siRNA (P < 0.001; n = 8) (Fig. 7A), demonstrating that hyperglycemia-induced TGase activation in the diabetic retina was mostly contributed by TGase2. We then analyzed the effect of TGase2 siRNA intravitreal injection on hyperglycemia-induced vascular leakage in the diabetic retina. Vascular leakage was apparent in the diabetic retinas injected with control siRNA compared with the normal retinas (P < 0.001, n = 8); vascular leakage was blocked in the retinas of the contralateral eyes injected with TGase2 siRNA(P < 0.01; n = 8) (Fig. 7B). TGase2 was expressed in the all layers of retina, with higher levels in the ganglion cell and inner plexiform layers, which was significantly suppressed by TGase2 siRNA (Fig. 7C). These results demonstrate that TGase2 plays a key role in hyperglycemia-induced vascular leakage in the diabetic retina.

To further confirm the role of TGase2 in vascular leakage in the diabetic retina, in vivo TGase activity and vascular leakage were determined in the retinas of C57BL/6 and TGase2^−/− mice at 2 weeks after injection (n = 8). TGase activity dramatically decreased in the retinas of nondiabetic TGase2^−/− mice compared with nondiabetic C57BL/6 mice (P < 0.001), and this decreased TGase activity was not changed by hyperglycemia in the diabetic TGase2^−/− mice (P = 0.468). As expected, TGase activity increased in diabetic C57BL/6 mice (P < 0.001), confirming that most TGase activation in the diabetic retina was contributed by TGase2. We then studied vascular leakage in the retinas of nondiabetic and diabetic TGase2^−/− mice (n = 8). Vascular leakage was not induced in the retinas of diabetic TGase2^−/− mice compared with the nondiabetic TGase2^−/− mice (P = 0.682; n = 8) (Fig. 8B and C), which is consistent with the in vivo experiments using intravitreal injection of TGase2 siRNA illustrated in Fig. 7. Vascular leakage was also undetectable in the retinas of diabetic TGase2^−/− mice at 4 weeks after injection (data not shown). Our results demonstrate that intracellular ROS-induced TGase2 activation plays a key role in VEGF-mediated vascular leakage in the diabetic retina.

Diabetic retinopathy is a major microvascular complication and is the leading cause of blindness in adults (1,2). Retinal vascular leakage in early stages of diabetic retinopathy is predominantly caused by VEGF-mediated stress fiber formation and adherens junction disassembly (7,10,11). However, the molecular mechanism by which VEGF induces vascular permeability in the diabetic retina is not clearly understood. Thus in this study we designed an in vivo TGase activity assay using confocal microscopy and explored the pivotal role of TGase2 in VEGF-induced vascular leakage in the retinas of diabetic mice. We demonstrated that hyperglycemia induced vascular leakage by activating TGase2 in diabetic retinas. VEGF activated TGase2 through sequential elevation of intracellular Ca²⁺ and ROS levels in endothelial cells. VEGF-induced stress fiber formation and adherens junction disruption was suppressed by the TGase inhibitors cystamine and MDC and by TGase2 siRNA. In vivo TGase2 activity was also elevated in the retinas of diabetic mice, but this was suppressed by intravitreal injection of TGase inhibitors. Additionally, intravitreal injection of mouse TGase2 siRNA prevented hyperglycemia-induced TGase activation and microvascular leakage in the diabetic retina. C-peptide, which inhibited VEGF-induced ROS generation and vascular leakage, prevented in vivo TGase2 activation in the diabetic retina. The ROS scavengers NAC and Trolox also prevented hyperglycemia-induced activation of TGase in the diabetic retina. The role of TGase2 in VEGF-induced vascular leakage was further demonstrated using diabetic TGase2^−/− mice. Thus ROS-mediated activation of TGase2 plays a key role in VEGF-induced vascular leakage by stimulating stress fiber formation and VE-cadherin disruption in the retinas of diabetic mice (Fig. 8D).

It is likely that TGase2 is a pivotal enzyme in the pathogenesis of diabetic complications. TGase2 is expressed in various ocular tissues, including the cornea, retina, trabecular meshwork, and lens (24,32); however, the role of TGase2 in diabetic retinopathy is not known. In this study we elucidated the essential role of TGase2 in VEGF-induced vascular leakage in the retinas of diabetic mice. Additionally, we recently reported that TGase2 is involved in diabetic vasculopathy resulting from apoptosis in endothelial cells and in the aorta, heart, and kidneys of diabetic mice (28). Another study using streptozotocin-induced diabetic rats demonstrated that TGase inhibition ameliorates the progression of diabetic nephropathy (33), suggesting the possible role of TGase2 in this disease. TGase2 is also associated with cataracts through cross-linking crystallins and vimentin (18,25). Thus TGase2 is likely the key enzyme for diabetic vascular permeability as well as diabetic vasculopathy and might be a potential therapeutic target for treating diabetic complications and ocular diseases.

Intravitreal injection of the TGase inhibitors MDC and cystamine or TGase2-specific siRNA prevented hyperglycemia-induced TGase2 activation and vascular leakage in the retinas of diabetic mice. Several low-molecular-weight inhibitors have been reported to prevent TGase2 transamidating activity, and the most widely used TGase inhibitor in vivo is cystamine (23). Cystamine is used for the treatment of corneal crystals in nephropathic cystinosis (34,35). Cystamine was also used in clinical trials of Huntington disease and cystic fibrosis (34,35). Cysteamine, the reduced form of cystamine, was recently reported to restore cystic fibrosis transmembrane conductance regulator function in patients with cystic fibrosis (36). The previous clinical or preclinical trials using cystamine to treat neurodegenerative diseases and cystic fibrosis suggest it would be possible to use cystamine for the prevention and treatment of diabetic vascular leakage. Moreover, TGase2-specific siRNA might be another target molecule to apply topically to the eyes for preventing diabetic retinopathy or other TGase2-associated ocular diseases.

In endothelial cells, adherens junctions comprise VE-cadherin and several protein partners, including β-catenin (37). VE-cadherin interacts directly with β-catenin, and VEGF induces adherens junction disruption by their tyrosine phosphorylation and subsequent dissociation (37). In ovarian cancer cells, TGase2 recruits c-Src, which in turn phosphorylates β-catenin and releases β-catenin from E-cadherin (31). Therefore we tested whether β-catenin and c-Src can participate in VEGF-induced adherens junction disruption and vascular leakage. In this study the TGase inhibitor cystamine inhibited VEGF-induced c-Src phosphorylation, β-catenin disassembly, and endothelial permeability in HRECs. Dasatinib, which blocked VEGF-induced c-Src phosphorylation, also prevented VEGF-induced β-catenin disassembly and endothelial permeability. These results suggest VEGF induces vascular leakage in endothelial cells through TGase2-mediated intracellular events involving c-Src phosphorylation, dissociation of β-catenin from VE-cadherin, and subsequent adherens junction disassembly.

TGase2 is localized in various subcellular compartments, such as the cytosol, nucleus, mitochondria, and extracellular space (13). It is likely that intracellular TGase2 participates in VEGF-induced vascular leakage because TGase2 is mainly located in the cytoplasm and nucleus and is activated by extracellular signals (13,38). Additionally, TGase2 regulates β-catenin signaling through c-Src phosphorylation in the intracellular organelles (31). However, whether extracellular TGase2 is involved in VEGF-induced vascular leakage is not clear. It was recently reported that extracellular TGase2 inhibits VEGF-induced angiogenesis by impeding the interaction of VEGF with heparan sulfate proteoglycans and subsequent VEGF receptor-2 signaling independent of its transamidating activity (39). There are, however, contradictory reports on the involvement of extracellular TGase2 activity in VEGF-mediated angiogenesis (40). Thus it is necessary to elucidate whether extracellular TGase2 participates in VEGF-induced endothelial permeability because TGase2 activation was essential for this pathological process.

Human C-peptide is a 31–amino acid peptide that is released from β-cells into the peripheral circulation in equimolar concentrations with insulin (1). C-peptide is beneficial in the prevention of diabetic complications including neuropathy, nephropathy, and inflammation (1,41). We recently demonstrated that C-peptide replacement therapy prevents diabetic vasculopathy, retinopathy, and impaired wound healing (11,28,30). C-peptide prevents diabetic vasculopathy by inhibiting ROS-mediated TGase2 activation and endothelial apoptosis in mouse aorta (28). C-peptide also prevents VEGF-mediated microvascular permeability by inhibiting ROS-mediated stress fiber formation and VE-cadherin disruption in diabetic mouse retina (11). In this study we demonstrated that intravitreal injection of C-peptide prevented ROS-mediated activation of TGase2 in the diabetic retina. Thus C-peptide can provide beneficial effects on diabetic vasculopathy and retinopathy by inhibiting intracellular events involving ROS and TGase2 (Fig. 8D). However, it is necessary to understand the function of and mechanism of action for intracellular ROS and TGase2 in diabetic neuropathy, nephropathy, and impaired wound healing.

In conclusion, we found that VEGF activated TGase2 by sequential elevation of intracellular Ca²⁺ and ROS levels and induced vascular integrity collapse through stress fiber formation and VE-cadherin disruption. ROS-mediated TGase2 activation played a pivotal role in the VEGF-induced collapse of vascular integrity in endothelial cells and hyperglycemia-induced microvascular leakage in the retina of diabetic mice. Thus TGase2 might be a potential therapeutic target for treating diabetic retinopathy or TGase2-associated ocular diseases.

Funding. This work was supported by the National Research Foundation of Korea (2013R1A2A1A09008193 and 2015R1A4A1038666).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.-J.L. and S.-H.J. researched data and wrote the manuscript. S.-H.K. and M.-S.K. researched data. S.L., J.Y.H., and S.-Y.K. contributed samples and to the discussion. Y.-M.K. contributed to the discussion. K.-S.H. designed research, analyzed data, contributed discussion, and wrote the manuscript. K.-S.H. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.