Abstract
purpose. Diabetic retinopathy increases the permeability of the blood–retinal barrier, but the specific vessels that become permeable have not been identified. Both transcellular and paracellular pathways of vascular solute flux have been proposed. This study was conducted to test the hypothesis that paracellular flux contributes to increased retinal vascular permeability after VEGF treatment or diabetes, and to map the types of vessels that became permeable.
methods. Regions of paracellular flux were identified by perfusion with fluorescent concanavalin A (ConA). Rats were injected intravitreally with VEGF or made diabetic with streptozotocin (STZ). After specified times, the rats were perfused with fixative followed by ConA, which binds to the basement membrane but not the luminal surface of endothelial cells. With this approach, ConA labels only blood vessels with paracellular permeability. Retinas were also labeled by immunofluorescence for the tight junction proteins occludin and claudin-5 and examined by confocal microscopy.
results. ConA labeling increased in the superficial arterioles and postcapillary venules, 2 weeks after the onset of diabetes. After 1 month, ConA labeling dramatically increased and extended to the capillaries of the outer plexiform layer. There was an inverse relationship between occludin immunoreactivity and ConA binding, but no change in claudin-5 immunoreactivity was detected. Injection of VEGF gave similar results.
conclusions. Diabetes and VEGF increase paracellular vascular permeability in the retina, associated with redistribution of occludin. This permeability begins in the superficial arterioles and postcapillary venules and progresses to the capillary bed.
The blood–retinal barrier, like the blood–brain barrier, is composed of continuous tight junctions between vascular endothelial cells. Increased retinal vascular permeability is a well-established consequence of diabetes,
1 2 causing macular edema, which correlates with vision loss in diabetic retinopathy.
3 The retinal vasculature becomes permeable to sodium fluorescein within 8 days after the onset of streptozotocin (STZ)-induced diabetes in rats,
4 whereas increased permeability to albumin has been measured after 5 weeks
5 and 3 months
6 of diabetes and also occurs in spontaneously diabetic BB/W rats.
7 Vascular endothelial growth factor (VEGF) can dramatically increase vascular permeability and is elevated by diabetes.
8 9 10 11 12 13 Conversely, injection of a soluble VEGF receptor reduces retinal vascular permeability,
14 demonstrating the importance of this cytokine in diabetic retinopathy.
The route of solute flux across the vascular endothelium has been debated for some time. Labeling of intracellular vesicles by horseradish peroxidase and albumin increases in spontaneously diabetic rats, suggesting transcellular flux.
15 16 17 This observation has been attributed to the formation of caveolae in response to VEGF
18 and to the formation of linked vesicles called vesiculovacuolar organelles.
19 20 Others propose that permeability may also include paracellular routes in which solutes diffuse between cells.
21 22 Paracellular permeability is likely to involve altered regulation of endothelial cell tight junction proteins.
23 24 STZ diabetes increases vascular permeability, decreases the content of the tight junction protein occludin,
6 and causes redistribution of occludin immunoreactivity from the plasma membrane to the cytoplasm.
25
In the present study, we directly assessed the contribution of retinal vascular paracellular permeability in response to diabetes and VEGF treatment by using the plant lectin concanavalin A (ConA).
26 ConA has specific binding affinity for α-
d-glucosyl and α-
d-mannosyl glycoproteins, which are common in the endothelial basement membrane. Thurston et al.
26 found that in the trachea of rats injected with substance P, ConA binding was absent in normal vessels but increased on the basolateral surface of inflamed venules and colocalized with monastral blue extravasation. We used ConA perfusion in STZ diabetic rats or after VEGF injection to identify regions of paracellular permeability in the retinal vasculature. We report that ConA labeling dramatically increased within 1 month of the induction of STZ diabetes and 30 minutes after intravitreal injection of VEGF. ConA binding was associated with redistribution of occludin immunoreactivity from the plasma membrane to the cell interior, but claudin-5 immunoreactivity appeared unaltered. The increase in permeability began in the superficial arterioles and postcapillary venules. Over time, the capillary network of the outer plexiform layer also became permeable. These data provide the first direct evidence that diabetes induces paracellular vascular permeability in the retina by redistribution of occludin.
STZ, phosphate-buffered saline (PBS), and Triton X-100, were purchased from Sigma-Aldrich (St. Louis, MO). Paraformaldehyde and bovine serum albumin (BSA) were from Fisher Scientific (Fair Lawn, NJ). Recombinant human VEGF-165 was from R&D Systems (Minneapolis, MN). Alexa Fluor 488–conjugated concanavalin A (ConA) was from Molecular Probes (Eugene, OR). The polyclonal rabbit anti-occludin antibody was from Zymed (South San Francisco, CA) and polyclonal anti-claudin-5 was made by Bethyl Laboratories (Montgomery, TX) against the peptide SAPRRPTANGDYDKKNYV, which corresponds to the C-terminal cytoplasmic domain of mouse claudin-5, as described in Liebner et al.
27 Donkey serum and secondary antibodies were from Jackson ImmunoResearch (West Grove, PA).
Male Sprague-Dawley rats weighing 150 to 175 g (Charles River Laboratories, Wilmington, MA) were housed in the Penn State College of Medicine animal facility, in accordance with the Institutional Animal Care and Use Committee guidelines. All animal experiments were performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and with the NIH Guidelines for the Care and Use of Laboratory Animals. All rats were housed in suspended wire-bottomed cages with ad libitum food and water, on a 12-hour light-dark schedule. Diabetes was induced by a single tail vein injection of STZ (65 mg/kg, 10 mM sodium citrate buffer, pH 4.5) and confirmed 3 days later by blood glucose higher than 250 mg/dL (Lifescan; Johnson and Johnson, Milpitas, CA). The weight and blood glucose of each rat was measured before death by decapitation under deep pentobarbital anesthesia.
For intravitreal injection of VEGF, Sprague-Dawley rats were anesthetized intramuscularly with a mixture of ketamine and xylazine (53.3 mg/kg ketamine, 5.3 mg/kg xylazine). After the corneal reflex was abolished, 5 μL of 0.1% BSA in PBS, containing 1 ng VEGF was injected into the vitreous chamber of one eye and vehicle was injected into the contralateral eye, with a syringe (Hamilton, Reno, NV) fitted with a 30-gauge needle. The rats were perfused 15, 30, or 60 minutes after intravitreal injections.
The ConA perfusion technique was adapted from one used to determine in vivo vascular permeability in lung.
26 Rats were deeply anesthetized with ketamine and xylazine. The descending aorta was clamped, and all solutions were circulated by transcardiac perfusion through an 18-gauge needle inserted into the left ventricle of the heart, using a linear flow pump (Harvard Apparatus, Holliston, MA). All perfusate exited through a hole placed in the right atrium. The flow rate was set at 10 mL/min for all solutions. This low perfusion rate was adequate to clear blood from the vasculature without increasing pressure to a point where the fine capillaries of the retina were ruptured. The perfusion solutions were, in order: 50 mL PBS and heparin (0.1 mg/mL); 50 mL of 2% paraformaldehyde; 50 mL PBS; 50 mL PBS with 1% BSA; 50 mL ConA (20 μg/mL) in PBS with 1% BSA; and 50 mL PBS containing 1% BSA. Retinas were dissected in ice-cold PBS, postfixed by immersion in fresh 2% paraformaldehyde for 5 minutes, and rinsed.
Whole retinas were blocked and permeabilized in 10% donkey serum with 0.3% Triton X-100 in PBS for 1 to 2 hours. The retinas were transferred to block solution with primary antibodies and incubated for 3 days at 4°C, as described previously.
25 The primary antibodies were rabbit anti-occludin (1:2000) and rabbit anti-claudin-5/6 (1:1000). The retinas were transferred to the secondary antibody for 24 hours at 4°C after washing in PBS with 0.3% Triton. The secondary antibody was rhodamine red-X-conjugated donkey anti-rabbit F(ab′)
2 (1:2000). Retinas were mounted flat on microscope slides by making four radial cuts, with the inner limiting membrane uppermost and covered in aqueous mounting medium (Poly/Aquamount; Polysciences Inc., Warrington, PA).
Digital confocal images (1024 × 1024 pixels) were captured with a confocal microscope (Nikon/Bio-Rad, Richmond, CA; or model TCS SP2 AOBS; Leica, Deerfield, IL, for the images in
Fig. 7 only). Digital images from experimental and control retinas were captured by using identical photomultiplier tube (PMT) gain settings and processed identically (Photoshop ver. 6.0.1; Adobe Systems, Mountain View, CA, for the Macintosh; Apple Computer, Cupertino, CA).
Statistical comparisons were made by χ2 test of independence with α set to 0.05 (Instat 3.0a for Macintosh; GraphPad Software, San Diego, CA).
To determine whether diabetes increases vascular paracellular flux in the retina, STZ-diabetic and control rats were perfused with fixative followed by the lectin ConA linked to a fluorescent marker. ConA binding sites are more abundant on the abluminal side of endothelial cells than on the luminal side, and significant labeling therefore occurs only in places where the basement membrane is directly accessible from the vascular lumen.
28 Under the conditions used here, active transport cannot occur because of prior fixation with paraformaldehyde, before perfusion with ConA. The wholemount retinas were observed by confocal microscopy at low magnification (10× objective), so that all the blood vessels of the retina were visible within one plane of focus (single confocal optical section). The retinas were always mounted with the inner limiting membrane facing up, so that the types of blood vessels could be identified by their caliber and point of focus within the tissue. In control retinas the superficial venules were labeled with ConA
(Fig. 1A) indicating that these vessels may be permeable under normal conditions. After 2 weeks of STZ diabetes, there was an increase in ConA labeling in the branches of large venules, postcapillary venules, and arterioles in the ganglion cell layer
(Fig. 1B) . After 4 weeks of STZ diabetes, the ConA labeling in arterioles was brighter and more extensive than in the control retinas and the 2-week diabetic retinas. ConA labeling also extended throughout the postcapillary venules and often into the capillary bed of the outer plexiform layer
(Fig. 1D) . Similar labeling was obtained in rats after 8 weeks of STZ diabetes
(Fig. 1F) . It is of particular note that labeling was evenly distributed in the inner retina, rather than being concentrated at foci, suggesting that changes occurred in all the blood vessels of a particular class, rather than at regionalized localities.
Normal rats were injected with 1 ng VEGF in the vitreous cavity. Paracellular permeability was examined by ConA perfusion, as described earlier. In vehicle-injected eyes, ConA labeling was again restricted to the large-caliber superficial blood vessels, primarily the venules
(Fig. 2A) . ConA labeling was increased in the postcapillary venules 15 minutes after injection of VEGF
(Fig. 2B) . The distribution of ConA fluorescence was more extensive 30 minutes after VEGF injection, labeling entire networks of postcapillary venules and arterioles
(Fig. 2C) . A similar pattern occurred 60 minutes after VEGF injection, and labeling included some capillaries
(Fig. 2D) .
To quantify the degree of ConA labeling in retinas from both STZ diabetic and VEGF-treated rats, photographs were taken with the 10× objective and independently assessed by three different investigators in a masked fashion. Each image was assigned a ranking score on a scale of 1 to 5: 1 for ConA labeling only in venules; 2 for labeling in all large vessels; 3 for labeling in large vessels and some postcapillary venules; 4 for labeling of large vessels, all postcapillary venules, and some capillaries; and 5 for labeling in all vessels, including capillaries. For each retina, the frequency of each ranking score (from 1 to 5) was expressed as a percentage of the total possible number of scores from the three investigators. The percentage frequency of each ranking score was plotted on a bar chart
(Fig. 3) . Data from the 1- and 2-month STZ diabetic groups were combined for a total of nine STZ diabetic and nine control rats. The photographs of STZ diabetic rat retinas were ranked significantly higher than those of control retinas (
P < 0.001, χ
2 test of independence,
Fig. 3A ). The photographs from VEGF-treated retinas were also ranked significantly higher than their paired controls (
P < 0.005,
Fig. 3B ). It is interesting to note that the ConA labeling after VEGF treatment was less intense than in the STZ diabetic rats. The slight difference in variation of the control group in the second experiment
(Fig. 3B) may be due to the intraocular vehicle injection.
Effect of Diabetes on Occludin Redistribution Associated with Paracellular Permeability
We have reported observation of regions of altered occludin immunoreactivity in the vasculature of rat retinas after 4 months of STZ diabetes.
25 To examine the relationship between tight junction proteins and paracellular permeability, fluorescence-labeled ConA was perfused in STZ diabetic rats, as described, and occludin was labeled by fluorescence immunohistochemistry. In control rats, ConA labeled only the venules and was absent in the arterioles and capillaries
(Fig. 4A) . In 2-week STZ diabetic rats the arterioles were also labeled with ConA fluorescence
(Fig. 4B) .
ConA labeling and occludin immunoreactivity were examined in rat retinas after 1 month of STZ diabetes
(Fig. 5) . Occludin immunoreactivity was present throughout the arterioles and capillaries of retinas from control rats, whereas there was no detectable ConA labeling in these vessels
(Figs. 5A 5B) . After 1 month of STZ diabetes, occludin immunoreactivity was reduced, and there was extensive ConA labeling in the superficial arterioles and venules as well as the capillary bed in the outer plexiform layer, identified by focusing through the tissue
(Figs. 5C 5D) . These data suggest that paracellular permeability increases in the arterioles, venules, and capillaries within 1 month of STZ diabetes.
To investigate further the relationship between occludin immunoreactivity and ConA labeling, higher magnification images of retinal vessels from 1-month STZ diabetic rats were examined
(Fig. 6) . Localized regions of positive occludin immunofluorescence corresponded to regions with less intense or completely absent ConA fluorescence, whereas vessels with low occludin immunofluorescence had more intense ConA fluorescence. These data suggest that occludin immunoreactivity at the plasma membrane is inversely related to ConA permeability and supports the hypothesis that paracellular permeability contributes to diabetic retinopathy.
Occludin immunoreactivity was redistributed and reduced in arterioles and postcapillary venules after VEGF injection
(Fig. 7) . Occludin immunoreactivity at the junctions between cells was intense in retinas from vehicle-injected eyes
(Fig. 7A) . Immunoreactivity was punctate and interrupted in vessels of retinas 15 minutes after VEGF injection
(Fig. 7B) . Immunoreactivity was dramatically reduced 60 minutes after VEGF injection
(Fig. 7C) . These data suggest that VEGF alters the distribution and reduces the immunoreactivity of occludin in a way similar to that in diabetes.
Retinas were labeled by fluorescence immunohistochemistry for claudin-5 in control eyes after intravitreal injection of VEGF. Claudin-5 immunofluorescence was similar in arterioles and venules of control retinas
(Figs. 8A 8B) and was unchanged 30 and 60 minutes after VEGF injection
(Figs. 8C 8D) . Claudin-5 immunoreactivity was also unchanged in retinas from rats after 1 month of STZ diabetes, even in regions with increased ConA labeling (data not shown). Therefore, ConA permeability was specifically associated with redistribution of immunoreactivity for occludin and not claudin-5 in response to VEGF injection and diabetes.
The purpose of this study was to map retinal blood vessels with increased paracellular permeability due to diabetes. A histologic technique that uses the plant lectin ConA was used to determine the spatial relationship between tight junction protein expression and localized increases in paracellular vascular permeability.
26 ConA binds to α-
d-mannosyl and α-
d-glucosyl glycoprotein residues, which are located in abundance on the abluminal side of brain and retinal vascular endothelial cells, but infrequently on the luminal side.
28 When AlexaFluor 488–conjugated ConA is perfused after fixation with 2% paraformaldehyde it binds to regions where the basement membrane is exposed to the vascular lumen, labeling regions of paracellular permeability. This perfusion technique was originally used to identify regions of increased paracellular permeability in the tracheal vasculature after induction of acute inflammation with substance P.
26 In that study, ConA binding corresponded to regions of plasma leakage identified by electron microscopy for monastral blue extravasation, which confirmed that ConA binding corresponded to paracellular permeability. Another study also described ConA binding associated with increased vascular permeability in the dystrophic rat model for proliferative retinopathy.
29
ConA binding was more extensive in retinas of STZ diabetic rats than in retinas of age-matched control animals. This study demonstrates that paracellular permeability increases in the retinas of STZ diabetic rats as early as 2 weeks after the induction of diabetes. The vessels most consistently having an increase in labeling were the superficial postcapillary venules and regions of arterioles. Some retinas had extensive labeling in the capillary bed of the outer plexiform layer after 1 month of STZ diabetes. These data show that paracellular permeability to a 52- to 104-kDa solute increased soon after the onset of diabetes in STZ diabetic rats, implying that significant gaps appeared between the vascular endothelial cells. In addition, vascular permeability occurred throughout the retina rather than at focal points. Other studies have described focal increases in vascular permeability to endogenous albumin in rat retinas and postmortem tissue after diabetes.
16 30 However, these studies examined albumin extravasation rather than the accessibility of the vascular basement membrane. Clinical studies tend to identify focal points of increased vascular leakage on the macro scale, but it is likely that generalized increases in permeability to water, giving rise to macular edema, are not so easily detected by fluorescein angiography. Therefore, the difference in findings may be due to the sensitivity of the ConA technique and the ability to view the entire retina using this approach, rather than sampling histologic sections.
There was an inverse relationship between occludin immunoreactivity and paracellular permeability. This relationship was demonstrated by the pattern of occludin immunoreactivity in normal retinas, which was high in the arterioles and capillaries and progressively decreased in venules with propinquity to the optic disc.
25 Furthermore, ConA labeling was only present in the venules of control retinas. On treatment with VEGF or diabetes, occludin immunoreactivity decreased at the plasma membrane with a concomitant increase in ConA labeling. Occludin protein content measured by Western blot was reduced by approximately 35% after 3 months of STZ diabetes.
6 Taken together, these data suggest that reduced occludin content and increased vascular permeability are directly related in diabetes. However, relocation of occludin from the plasma membrane to the cytoplasm may be the mechanism responsible for the onset of permeability, because VEGF did not reduce total occludin content in rat retinas over a similar time course, as observed in Western blot analysis,
31 but dramatically reduced occludin immunoreactivity at the cell borders. The possibility that the changes in occludin immunoreactivity occur in response to increased vascular permeability cannot be ruled out, but this causal relationship seems less attractive because of the absence of a likely mechanism. The subcellular redistribution of occludin protein was noted previously in the vasculature of 1-month STZ diabetic rat retinas.
25 Redistribution of occludin and ZO-1 also occurs in cultured brain microvascular endothelial cells in response to VEGF
32 and hypoxia.
33 In addition, platelet-derived growth factor, also induces tight junction protein redistribution and increased permeability in Madin-Darby canine kidney cells.
34 These data are consistent with the hypothesis that occludin migrates into a different cellular compartment after stimulation with growth factor, allowing paracellular permeability.
34 Therefore, redistribution of tight junction proteins from the plasma membrane into the cytoplasm may be part of the mechanism that regulates paracellular permeability at the blood–retinal barrier.
The distribution of claudin-5 immunofluorescence was not affected by diabetes or VEGF. Claudin-5, or transmembrane protein deleted in velocardiofacial syndrome (TMVCF)
35 36 is ubiquitously expressed by endothelial cells.
37 In the current work, we found that claudin-5 immunofluorescence did not change in response to diabetes or VEGF injection, demonstrating that occludin redistribution is a specific response. However, these data do not exclude a role for claudin-5 in retinal vascular permeability. Similar conclusions have been made in other studies of claudins. Hypoxia induced a 2.6-fold increase in permeability to sucrose in bovine brain microvessel endothelial cells, accompanied by altered occludin, ZO-1, and ZO-2 immunoreactivity, but the protein expression and distribution of claudin-1 remained unchanged.
38 Brain vascular claudin-1 expression was also unchanged in an inflammatory pain model, despite an increase in brain vascular permeability and associated changes in ZO-1.
39 Morcos et al.
40 described the distribution of claudin-1 in the retina; however, our study is the first demonstration of immunoreactivity for the endothelial cell–specific claudin-5 in the retina.
In conclusion, this is the first study to provide direct identification of regions of paracellular vascular permeability in the retina and, to our knowledge, is the first study to map the vessels with increased paracellular permeability due to diabetes. Increased permeability begins with the arterioles and postcapillary venules of the superficial vasculature followed later by the capillary network in the outer plexiform layer. These data suggest that redistribution of the transmembrane tight junction protein occludin from the plasma membrane to the cell interior regulates paracellular permeability. Future studies of the molecular mechanism of occludin redistribution will shed light on the cellular events that regulate permeability of the blood–retinal barrier.
The authors thank Thomas W. Gardner for useful comments during preparation of the manuscript.