Proteolytic Degradation of VE-Cadherin Alters the Blood-Retinal Barrier in Diabetes

Brown Norway rats were purchased from Harlan Laboratories. The glycohemoglobin assay kit was purchased from Stanbio. Bovine retinal microvascular endothelial cells and MCDB 131 complete media were purchased from VEC Technologies. Fibronectin-coated dishes were obtained from Becton Dickinson. Glycoaldehyde-modified AGE-BSA was purchased from Calbiochem. Rabbit polyclonal antibody to VE-cadherin was purchased from Alexis. Purified MMP-2 and MMP-9 were purchased from R&D Systems. The metalloproteinase inhibitor BB-94 was provided by British Biotech. All other reagents were purchased from Sigma-Aldrich, unless otherwise noted.

Brown Norway rats were injected with a single intraperitoneal injection of streptozotocin (60 mg/kg i.p.) in 10 mmol/l citrate buffer, pH 4.5. Control nondiabetic rats received injections of an equal volume of citrate buffer only. Animals with plasma glucose concentrations >250 mg/dl, 24–48 h after streptozotocin injection, were considered diabetic and included in these studies. Body weight was measured twice each week, and insulin (0.5 IU) was given based on the stability of body weight. The tolerated maximum difference in weight between diabetic and nondiabetic rats was 40%. Total glycated hemoglobin (GHb) was measured in each animal at the end of the 2- or 8-week period. All experiments were consistent with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and were done in accordance with institutional animal care and use guidelines.

Experiments were performed using bovine retinal microvessel endothelial cells (BRMVECs). The endothelial cells were grown in MCDB-131 media containing 10% fetal bovine serum, endothelial cell growth supplement, and antibiotics. Passages 3–8 were used for all experiments.

Vascular permeability in the retina was measured following 2 and 8 weeks of diabetes using an Evans blue quantitation technique (25). Briefly, rats were anesthetized and received an intravenous injection of 45 mg/kg Evans blue dye. Two hours later, the animal was reanesthetized, a 0.3-cc blood sample was obtained, and the animal was perfused via the left ventricle with PBS followed by 1% paraformaldehyde. The retinas were collected in PBS, dried for 2 h, and weighed. The Evans blue dye was extracted from the dried retinas with formamide at 70°C for 18 h and centrifuged through a 30,000-MW filter (Millipore, Bedford, MA). The absorbance of the retinal extract and a 1:1,000 dilution of plasma was measured by spectrophotometry at 620 and 740 nm. The concentration of dye in the extract and plasma was calculated from a standard curve of Evans blue dye. The BRB permeability was calculated as follows, and data were expressed as μl · mg⁻¹ · h⁻¹: Evans blue retina (ng)/Evans blue plasma (ng/μl) × circulation time × retina dry weight (mg).

In some experiments, normal nondiabetic rats were anesthetized and microinjected intraocularly with 5 μg VE-cadherin antibody or PBS in 1 μl (three eyes each from three different animals). Sham animals received a needle puncture only (four eyes from two different animals). Permeability was assessed 2 h later.

Retinal vasculature was selectively isolated from other components of the retina using a procedure previously described (26). Briefly, retinas were incubated in ice-cold sterile water for 1 h at 4°C followed by incubation with RQ DNase for 5 min. Microvascular networks were transferred to ice-cold water and cleaned of debris by gentle pipetting with a wide-bore pasteur pipette. Total mRNA was extracted from the microvascular networks using Trizol reagent (Invitrogen, Carlsbad, CA), and the integrity was confirmed by agarose gel electrophoresis. cDNA was synthesized using the TaqMan Reverse-Transcriptase reagents (Applied Biosystems, Foster City, CA). The following TaqMan assays for real-time RT-PCR were obtained from Applied Biosystems: VE-cadherin, RatCad5-RC50; MMP-2, Rn02532334; MMP-9, Rn00579162; and 18S, Hs99999901. Amplification and detection was performed using the Applied Biosystems 7500 Fast system. Data were derived using the comparative C_t method for triplicate reactions (27).

Endothelial cells were incubated in serum-free media for 24 h with AGE-BSA (50 μg/ml) with or without the metalloproteinase inhibitor BB-94 (2 μmol/l). Additional cultures were incubated for 2 h in serum-free media with either purified MMP-2 (0.1 μg/ml) or MMP-9 (0.1 μg/ml). The conditioned media from the various treatments were concentrated using centricon filters (30 kDa cutoff) and the total protein quantitated using the MicroBCA assay (Pierce). Equal amounts of total protein were loaded onto a 10% SDS-PAGE.

Proteins were transferred onto polyvinylidine fluoride membrane and probed with an antibody to VE-cadherin. The bands were visualized using a chemiluminescent detection kit (Pierce, Rockford, IL) and the GNOME-GENESNAP advanced image acquisition software (Syngene, Frederick, MD).

Levels and degree of activation of MMP-2 and MMP-9 were examined in cells by zymography. The conditioned media from cells cultured in serum-free media were collected after 48 h, and aliquots were removed for total protein determinations. Conditioned medium was subjected to electrophoresis in 10% polyacrylamide mini-gels, into which gelatin (1 mg/ml) was cross-linked. Zones of clearing corresponding to the presence of proteinases in the samples were quantitated using the Alpha Innotech Image analysis system.

BRMVECs were seeded onto fibronectin-coated Transwell inserts (0.4-μm pore size) and grown in complete media for 5 days. The media was changed every alternate day in the upper chamber until confluency was reached. The functional confluency of the cells was determined by measuring the TER (transendothelial resistance). Confluent BRMVECs exhibited a TER reading of 220 ± 10 ohms, which reflected a functional barrier in these cells. The cells were changed to serum-free media, and 5 μl of 40 mg/ml fluorescein isothiocyanate (FITC)-dextran (40,000 MW) was added to the top chamber. Different wells received AGE-BSA (50 μg/ml) with or without BB-94 (2 μmol/l). Fifty-microliter aliquots of media were removed from the bottom chamber at 0, 4, 24, and 48 h and analyzed by quantitative fluorimetry at 485 nm. In some experiments, cells were incubated with purified MMP-9 (0.1 μg/ml) and the media collected at 0, 2, and 4 h. Rates of permeability were calculated and the data expressed as fluorescence units per hour.

BRMVECs were passed onto glass coverslips coated with fibronectin and allowed to reach confluence. Cells were stimulated for 48 h with AGE-BSA (50 μg/ml) with or without BB-94 inhibitor (2 μmol/l). The cells were fixed with 4% paraformaldehyde for 5 min and stained with an anti–VE-cadherin antibody. The cells were examined using a fluorescence microscope, and images were captured and pseudo-colored using the MetaMorph Image Analysis program.

Microvessels in the retinas of 2- or 8-week-old diabetic and control animals were stained for VE-cadherin in retinal whole mounts. Briefly, retinas were isolated and fixed in 100% ethanol, delipidated in 100% ice-cold acetone, and permeabilized with 0.3% Triton X-100. Samples were incubated with an antibody to VE-cadherin at 4°C and washed extensively. VE-cadherin was localized using an FITC-labeled secondary antibody and Z-stack images obtained using a confocal microscope (LSM 510 META; Carl Zeiss Meditec, Gottingen, Germany) and analyzed qualitatively for differences in VE-cadherin staining. Analysis was limited to the mid-peripheral region of the retina and up to 12 separate areas per retina photographed.

The expression of VE-cadherin on the cell surface was determined by a cellular enzyme-linked immunosorbent assay (ELISA). Confluent monolayers of BRMVEC on 24-well plates were untreated or treated with purified MMP-2 (0.1 μg/ml for 2 h), MMP-9 (0.1 μg/ml for 2 h), AGE-BSA (50 μg/ml for 24 h), or AGE-BSA plus BB-94 inhibitor (2 μmol/l for 24 h). After treatment, the cells were fixed with 0.025% glutaraldehyde for 5 min and washed extensively with PBS. Cells were incubated overnight at 4°C with a VE-cadherin antibody (2.5 μg/ml). Cells were washed extensively and incubated for 1 h at room temperature with a horseradish peroxidase–labeled anti-rabbit antibody. The cells were washed, and the level of VE-cadherin on the cell surface was colorimetrically estimated following the addition of o-phenylenediamine dihydrochloride substrate and measuring the absorbance at 492 nm. Controls included cells incubated with secondary antibody only and cells incubated with the colorimetric substrate only.

For all quantitative experiments, statistical analyses of data were performed using either an unpaired t test or a one-way ANOVA using the Prism4 software (GraphPad Software).

Rats were injected with streptozotocin to induce diabetes and were used 2 or 8 weeks later for these studies. Animals with blood glucose levels >250 mg/dl were considered diabetic and were included in the study. The blood glucose levels for the nondiabetic animals averaged 167 ± 17 mg/dl and animals with diabetes 450 ± 44 mg/dl. The ratio of glycated hemoglobin to total hemoglobin in blood samples was elevated almost twofold in diabetic animals compared with nondiabetic controls at the end of both 2 and 8 weeks.

The degree of retinal vascular permeability was determined using Evans blue dye, a technique that effectively measures the movement of albumin from the plasma into the surrounding tissues. Retinal vascular permeability was significantly increased in the diabetic animals (0.055 ± 0.0034 μl · mg⁻¹ · h⁻¹) compared with the nondiabetic controls (0.030 ± 0.0035 μl · mg⁻¹ · h⁻¹) following 2 weeks of hyperglycemia (P = 0.0002). Similarly, at 8 weeks of diabetes, the permeability of the retinal vasculature in the diabetic animals (0.067 ± 0.0031 μl · mg⁻¹ · h⁻¹) was significantly increased compared with the nondiabetic controls (0.037 ± 0.0027 μl · mg⁻¹ · h⁻¹) (P = 0.0013).

The presence and distribution of VE-cadherin in the retinal microvasculature of control and diabetic rats were evaluated by immunohistochemistry and confocal microscopy. In nondiabetic rat retinas, the small (8- to 10-μm diameter) retinal vessels were positive for VE-cadherin staining (Fig. 1A and B). VE-cadherin was distributed along the endothelial intercellular junctions with no obvious areas of discontinuity. In contrast, after only 2 weeks of diabetes, the small vessels of the diabetic retinas demonstrated a substantial loss of VE-cadherin staining (Fig. 1C). The loss of VE-cadherin in the microvessels persisted following 8 weeks of diabetes (Fig. 1D). The decreased expression of VE-cadherin protein in the diabetic retina at 2 weeks was also associated with a significant downregulation in the level of VE-cadherin mRNA in isolated retinal microvessels (Fig. 1E and F).

To determine whether disruption of VE-cadherin alone alters retinal vascular permeability, normal nondiabetic rats received a single intravitreal injection of anti–VE-cadherin antibody (5 μg/μl) and were analyzed 2 h later for changes in vascular permeability using the Evans blue dye technique. The antibody recognizes the VE-cadherin ectodomain (AA1–258) and causes a disruption of VE-cadherin engagement in a confluent monolayer of bovine retinal endothelial cells, resulting in cell-cell separation (Fig. 2A and B). A nearly fivefold increase in retinal vascular permeability was detected in the antibody-injected eyes compared with the vehicle-treated or sham-injected eyes, suggesting a crucial role for VE-cadherin in maintaining BRB integrity (Fig. 2C).

We next examined the expression of MMPs in the retinal vasculature to determine whether the loss of VE-cadherin protein and increased retinal vascular permeability might, in part, be due to proteolytic turnover. We have previously shown that MMPs are upregulated in the retinas of animals after 12 weeks of diabetes (17). In animals with only 2 weeks of diabetes, the levels of MMP-2 and MMP-9 mRNA in isolated retinal vessels were also significantly increased compared with nondiabetic controls (Fig. 3).

To confirm a role for MMPs in disruption of BRB function, we assessed the ability of a MMP inhibitor to block the increases in retinal vascular permeability after 2 weeks of diabetes. The broad spectrum inhibitor BB-94 has been shown to effectively inhibit the MMP-induced retinal neovascularization in an animal model of oxygen-induced retinopathy (15). After daily intraperitoneal administration of BB-94 (1 mg/kg) to diabetic animals, the degree of retinal vascular permeability was nearly twofold lower compared with diabetic animals receiving PBS injections only (Fig. 4A). This functional change in permeability, in response to MMP inhibition, coincided with a structural change in the interendothelial junctions of the microvessels. The inhibition of MMPs prevented the loss of VE-cadherin from the cell surface of adjacent endothelial cells. VE-cadherin was present in a continuous uninterrupted pattern in the inhibitor-treated diabetic retinas similar to that seen in the nondiabetic animals (compare Fig. 4B with Fig. 1A and B).

The previous results suggest that during the development of diabetic retinopathy, hyperglycemia creates an environment that facilitates increased microvascular permeability associated with the upregulation of MMP expression and the loss of VE-cadherin. Several biochemical mechanisms have been proposed to explain the changes seen in the diabetic retina, including the formation and activity of AGE. The role of AGE in the regulation of endothelial barrier function was assessed using isolated bovine retinal microvascular endothelial cells. Cultured endothelial cells incubated with 50 μg/ml AGE-BSA demonstrated a significant increase in the rate of monolayer permeability compared with untreated cells (Fig. 5). This effect was found to be mediated by MMPs, as simultaneous treatment of cells with AGE-BSA and an inhibitor for MMPs resulted in a reduction of permeability toward control untreated levels. AGE stimulation could also account for the alterations in proteinase expression seen in the vasculature of the diabetic retina. The conditioned media of BRMVECs treated with AGE-BSA was analyzed by zymography and demonstrated a marked increase in both the pro and active forms of MMP-2 and MMP-9 after 48 h of stimulation (Fig. 6).

We next examined whether alterations in monolayer permeability observed in AGE-treated endothelial cells coincided with a change in VE-cadherin distribution. Control untreated cells displayed a continuous pattern of VE-cadherin staining associated with the lateral cell borders (Fig. 7A). Treatment with AGE-BSA caused a disruption in the pattern of VE-cadherin staining on the cell surface (Fig. 7B). The VE-cadherin staining was found to be punctate as opposed to continuous and was absent from regions where cell-cell contact was lost. Cells treated simultaneously with AGE-BSA and the MMP inhibitor demonstrated cellular morphology and VE-cadherin staining similar to untreated cells (Fig. 7C). This result suggests a possible mechanism by which AGE-BSA might alter adherens junctions in endothelial cells through the upregulation and activation of MMPs, leading to increased vascular permeability.

Retinal endothelial cells treated with AGE-BSA or purified MMPs release the VE-cadherin ectodomain while undergoing cell-cell separation. The ectodomain fragment of VE-cadherin was found in the conditioned media with a molecular weight of ∼75 kDa. The shedding of this portion of the protein appears to be a normal part of cadherin turnover in cells. This was demonstrated by the appearance of the fragment in media collected from cells with no treatment (Fig. 8A). Cells treated with purified MMP-9 or AGE-BSA demonstrated a significant increase in the amount of 75 kDa fragment released into the media. Cells treated with purified MMP-2 did not show the same degree of shedding, suggesting a degree of specificity to the process. The addition of BB-94 (2 μmol/l) to the AGE-BSA–treated cells resulted in a significant decrease in the level of shedding comparable with control conditions. The release of the VE-cadherin ectodomain into the media was consistent with a loss of VE-cadherin from the cell surface as determined by cellular ELISA (Fig. 8B). These results suggest that AGE causes a significant shedding of the VE-cadherin ectodomain through a specific MMP-dependent mechanism, which may lead to increased vascular permeability.

Diabetic retinopathy is one of the most common and most severe complications of diabetes, affecting ∼50% of all patients with diabetes (28,29). Breakdown of the BRB is an early feature of diabetic retinopathy and results in vascular leakage and the development of macular edema, which can cause distortion and loss of central vision (30–32). The molecular mechanisms that regulate BRB remodeling during diabetes remain unclear. The data presented here suggest that AGE stimulation results in the upregulation of proteinases that cleave VE-cadherin from the retinal endothelial cell surface, leading to increased vascular permeability.

In the present study, we have observed a significant decrease in the degree of VE-cadherin staining of the retinal vasculature coincident with an increase in vascular permeability. This was associated with a significant downregulation of VE-cadherin mRNA in isolated microvessels from the diabetic animals. A similar finding of decreased VE-cadherin has been reported in the retina of a human patient with nonproliferative diabetic retinopathy. Interestingly, this study also demonstrated that other junctional proteins involved in the formation and function of the BRB, including occludin and ZO-1, showed normal patterns of distribution (33). The present study is the first report demonstrating the loss of VE-cadherin and altered BRB function during early stages of diabetic retinopathy. Decreases in occludin levels have been reported in the diabetic rat retina following 3 or 4 months of diabetes (34,35). Whether the distribution of occludin and other tight junction proteins are altered at 2 weeks of diabetes is unknown.

The importance of VE-cadherin in maintenance of the BRB was confirmed in this study as vascular leakage was increased nearly fivefold 2 h following a single intravitreal injection of a VE-cadherin disrupting antibody. This dramatic increase in permeability upon disrupting VE-cadherin alone involves a major alteration of the BRB including the apparent disruption of tight junction proteins. Cadherins, including VE-cadherin, may interact with occludin and/or claudin proteins through specific intracellular signaling molecules to regulate the function of these proteins and the interendothelial barrier (36). Previous studies have also reported that VE-cadherin is not only important for regulating interendothelial cell permeability but also for cell survival and angiogenesis, which may involve similar mechanisms (6,37,38).

The turnover or remodeling of VE-cadherin in interendothelial junctions may involve different mechanisms depending on the particular stimulus. A recent study describes a mechanism by which vascular endothelial growth factor and its receptor induces the rapid internalization and recycling of VE-cadherin (39). This mechanism may predominate in situations requiring a rapid and reversible alteration of vascular permeability. Proteolytic modification of VE-cadherin on the cell surface, as seen in the retina during diabetes, may represent an alternative mechanism initiated in response to a sustained pathological stimulus. Previous studies have demonstrated that E- and N-cadherin can be cleaved by MMPs in a variety of circumstances involving pathological vascular remodeling (40–44). In the present study, we have reported that the incubation of cultured retinal microvascular endothelial cells with purified MMPs results in an MMP-specific shedding of the VE-cadherin ectodomain and loss of cell-cell contact.

The VE-cadherin molecule has four highly conserved cadherin repeat domains that are used to negotiate cell-cell adhesion. MMP-9 is known to cleave specific peptide bonds such as Gly-Val, Gly-Leu, Gly-Asn, and Gly-Ser (45). The VE-cadherin extracellular repeats have one Gly-Leu residue and five Gly-Ser residues within amino acids 51–576. It is possible that MMP-9 can cleave VE-cadherin at one or more of these sites, resulting in shedding of parts of or the entire ectodomain. A detailed mapping of the proteolytic fragments using purified VE-cadherin is necessary to confirm the exact sites of cleavage.

Hyperglycemia and other diabetic metabolites, including AGEs, are known to have negative effects on the retinal microvasculature. Numerous studies have localized AGEs to the retinal microvessels and basement membrane in both humans and rats (46). These factors can also affect MMP expression and activity (17,47,48). In this study, we examined the role of AGE in MMP upregulation by vascular endothelial cells and in alteration of VE-cadherin expression. Incubation of cells with AGE-BSA resulted in the upregulation of MMPs, caused the shedding of VE-cadherin in an MMP-dependent manner, and increased monolayer permeability. The mechanism by which AGE upregulates MMPs is currently unknown. One hypothesis is that AGE interacts with its receptor RAGE on the endothelial cell surface and directly induces the expression of MMPs through the activation of nuclear factor-kB or other signaling molecules. Another possibility is the indirect upregulation of proteinases by AGE through increased growth factor synthesis and activity (49).

Data from this study confirm an important role for specific extracellular proteinases in alteration of the BRB during the development of diabetic retinopathy. The ability of a broad-spectrum MMP inhibitor to block retinal permeability changes substantiates the direct involvement of MMPs in the breakdown of the BRB. This result suggests a potentially useful alternative therapeutic approach to the treatment of macular edema in diabetic retinopathy.

We acknowledge the funding provided by the National Institutes of Health Grant EY12604 and Juvenile Diabetes Research Foundation Grant 1-2006-36.