Abstract

Homeostasis in the retina microenvironment is maintained by the proper function of the blood-retinal barrier (BRB), which regulates the movement of chemicals and cells between the intravascular compartment and the retina. The BRB consists of two major topographically distinct components: the endothelium of the retinal vessels (inner BRB) and the retinal pigment epithelium (outer BRB). The barrier function of the retinal vascular endothelium depends on its lack of fenestrations, whereas the ability of the retinal pigment epithelium to regulate solute transport depends on the apical tight junctions. The tight junctions are membrane fusion areas between adjacent cells that serve as a diffusion barrier for paracellular transport and as a ‘molecular fence’, restricting the free movement of transmembrane proteins, and thus maintaining cell polarity and the asymmetric distribution of transmembrane proteins. Among the most important proteins that are associated with tight junctions are occludin, zonula occludens and claudins. Pathologic increase in blood retinal permeability can be caused by endothelial or pericyte cell death, tight junction disassembly, or cytokines such as vascular endothelial growth factor. Several assays have been developed to allow detection, quantification and monitoring of BRB breakdown in experimental and clinical settings. Assays used in animal models include the injection of chromophores, such as Evans blue, horseradish peroxidase, and fluorescein; the imaging techniques include electron microscopy and MRI. In humans, fluorescein angiog-

raphy, vitreous fluorophotometry and optical coherence tomography are most commonly used. The disruption of the BRB contributes to the pathophysiology of several retinal diseases such as diabetic retinopathy, agerelated macula degeneration, retinopathy of prematurity, central serous chorioretinopathy, vascular occlusive and inflammatory diseases. Several medical and surgical treatments have been developed to restore normal BRB function. Traditional procedures such as laser photocoagulation and corticosteroids have been recently supplemented with vascular endothelial growth factor pathway inhibitors, anti-TNF-α agents, mammalian target of rapamycin inhibitors and PKCβ inhibitors. Early results from clinical trials offer hope for effective visionpreserving therapies.

Copyright © 2010 S. Karger AG, Basel

Although the mammalian retina is constantly exposed to the rich choroidal circulation, it maintains a high level of electrolyte and metabolite balance that is crucial for the proper retinal function and ultimately vision. This homeostasis is maintained by the proper function of the blood-retinal barrier (BRB) that regulates the transport of cells and chemical substances from the circulation to the retina, therefore the retinal microenvironment. The molecular basis for the BRB are tight junctions (TJs) between endothelial cells in the inner retina,

and between pigmented epithelial cells in the outer retina. The disruption of the BRB in the retinal vasculature or in neovessels underlies the pathophysiology of a variety of vision-threatening diseases of the retina. Restoration of the vascular stability and integrity improves visual outcomes and is currently a therapeutic goal for many ocular conditions.

Physiology of the Retinal Vascular Network

The retina is a highly specialized neural tissue that consists of seven layers: the nerve fiber layer, the ganglion cell layer, the inner plexiform layer, the inner nuclear layer, the outer plexiform layer, the outer nuclear layer and the photoreceptors (rods and cones). The majority of the retina blood supply (85%) is derived from the choroidal blood vessels, whereas the central retinal artery provides the remaining 15%. The central retinal artery gives out four main vessels as it runs through the optic nerve head and supplies three capillary networks: the radial peripapillary, the inner and the outer network. The most superficial capillary network is the radial peripapillary one, which runs in the inner part of the nerve fiber layer along the major arterial arcades. The inner capillary network runs in the ganglion cell layer, whereas the outer capillary network runs throughout the inner nuclear layer. The three networks form multiple anastomoses between them. The retinal area responsible for central vision is located in the center of the macula, called the fovea; it is avascular and the retinal vessels arc around it. The choroidal vasculature consists of fan-shaped lobules of capillaries derived from the long and short posterior ciliary arteries and from branches of the peripapillary arterial network.

Physiology of the Blood-Retinal Barrier

The BRB maintains a constant milieu by regulating the exchange of water, nutrients, metabolites, proteins and neurotransmitters, and the efflux

of toxic byproducts of metabolism. Moreover, it shields the neural retina from the circulating blood by restricting the entry of toxins, inflammatory cytokines, antibodies and circulating immune cells. The concept of the existence of the blood-tissue barrier in neural tissues was first introduced in the literature in 1885 by Goodman who demonstrated that trypan blue injected intravenously in the rat stained all tissues except the brain [1, 2]. In 1965, Ashton and Cuhna-Vaz demonstrated that intravenously injected histamine increased the vascular permeability of various ocular tissues except the retina [3], leading to the concept of the BRB [2]. Subsequent morphological studies showed that the retinal endothelial cells demonstrate an epithelial-like structure with ‘zonnulae occludentes’ between them.

Maurice and Cunha-Vaz performed morphological studies and permeability measurements and proposed that the BRB consists of two major components: the endothelium of the retinal vessels (inner BRB) and the retinal pigment epithelium (RPE; outer BRB) [2]. These two components are topographically distinct (the former is responsible for BRB functions in the inner retina, whereas the latter for the outer retina) and mechanistically independent. Therefore, it should be emphasized that the two different yet parallel sources of perfusion in the retina (the choroidal blood vessels and the central retina artery) are dependent on different mechanisms of the BRB: the endothelial cells of the choroidal capillaries have fenestrations similar to those of endothelial cells elsewhere in the body and rely entirely on the adjacent RPEs for BRB functions. In contrast, the endothelial cells of the retinal network capillaries lack fenestrations and exhibit all the specialized barrier properties of the BRB, while their surrounding pericytes, which contribute to a second line of defense in the blood-brain barrier, are approximately four times as numerous in the retina as in the brain [4].

There are no diffusional barriers between the extracellular fluid of the retina and the adjacent

vitreous, and the vitreous body does not hinder significantly the diffusion of solutes. It should be emphasized that not all aspects of the physiology of BRB have been well studied in a retina-related model. Several conclusions are derived from extrapolation based on observations in other natural barriers, such as the blood-brain barrier.

Molecular Biology of the Blood-Retinal

Barrier

The main routes used by water, solutes and proteins to move across endothelial and epithelial cell layers can be classified as transcellular vs. paracellular flux. Transcellular (transfer across the cell) can be via passive diffusion, facilitated diffusion (channel-facilitated transport), active transport (receptor-mediated uptake), endocytosis/pinocytosis (membrane invaginations across the cell surface that pinch off to form vesicles that move to the cell interior and are released on the other side, allowing nonspecific transport of material), and finally via pores or fenestrations. It should be noted that RPE cells and endothelial cells in the BBB and BRB lack fenestrations [5] and have profoundly decreased pinocytosis activity, while the choriocapillaris is fenestrated [6]. It is possible that the choriocapillaris endothelial cell fenestrations are regulated by vascular endothelial growth factor (VEGF), as intravitreal injection of the anti-VEGF antibody bevacizumab in cynomolgus monkeys significantly reduced these fenestrations, an effect that may be of clinical relevance in the treatment of macular edema [7]. Because the choriocapillaris is fenestrated, it is the RPE cells that form the outer BRB and regulate the environment of the outer retina. Like all epithelia and endothelia, the ability of RPE to regulate transepithelial transport depends upon two properties: apical TJs to resist diffusion through the paracellular spaces of the monolayer, and an asymmetric distribution of proteins to regulate vectorial transport across

the monolayer [8]. During development, these properties form gradually. Initially, the TJs are leaky, and the RPE exhibits only partial polarity. As the neural retina and choriocapillaris develop, there are progressive changes in the composition of the apical junctional complexes, the expression of cell adhesion proteins, and the distribution of membrane and cytoskeletal proteins [8]. Another aspect of RPE function is the active transport of water out of the retina into the choriocapillaris. This flow of water out of the retina helps maintain retinal attachment.

In addition to controlling the influx of solutes, the BRB also actively transports potentially noxious compounds out of the retina in order to maintain the ideal microenvironment for its function. Lactate is actively transported from the RPE cells to the choroid [9, 10]. The P-glycoprotein is present in the BRB and actively pumping lipophilic toxins and drugs out of the endothelial or RPE cell, back to the bloodstream [5, 11–13].

The paracellular flux (transfer between cells) is primarily regulated by the permeability of TJs. In pathologic situations, disassembly of the TJs and large gaps in the cellular continuum allow for breakdown of the barrier.

Tight Junctions

TJs are areas of apparent fusions between the closely apposed outer leaflets of plasma membranes of adjacent cells (endothelial and epithelial), where the intercellular space disappears forming continuous seals circling around the cell’s circumference like a belt. TJs serve as a highly selective diffusion barrier and strictly control the paracellular flux of water and solutes [14], allowing the separation of fluids on either side that have a different chemical composition. They also function as a ‘molecular fence’ restricting the free movement of integral cell membrane proteins, thus maintaining a different protein composition between apical and basolateral membrane,

which contributes to cell polarity. Over 40 proteins have been found to be associated with TJs, including transmembrane, scaffolding, and intracellular signaling proteins [14, 15], such as occludin, the zonula occludens (ZO) proteins, claudins, and others.

The link between BRB, TJ molecules and angiogenesis is a subject of intense investigation. In human placenta, junctional complexes regulate angiogenesis and vascular remodeling. According to Leach [16], there are two types of junctional adhesion phenotypes that are regulated by the differential expression of VEGF and angiopoietins 1 and 2. The ‘activated’ type has low immunoreactivity for TJ molecules such as occludin and claudin, and is found in highly angiogenic terminal capillaries, whereas the ‘tight’ type has high levels of these molecules and is found in quiescent capillaries [16].

Transmembrane Tight Junction Proteins

Occludin

Occludin, a 65-kDa protein, was the first transmembrane TJ protein discovered [17], and is present in TJs of both epithelial and endothelial cells. It has four transmembrane helices, a short intracellular loop, two extracellular loops, and 2 intracellular tails. The intracellular N-terminal cytoplasmic tail interacts with the E3 ubiquit- in-protein ligase Itch, resulting in the ubiquitination of occludin, which promotes its degradation by the proteasome [18]. Cyclic AMP promotes disassembly of the TJs by promoting proteasome-mediated degradation of occludin [19]. This pathway provides a mechanism for cytokine-induced regulation of TJ function. The two extracellular loops can bind to the occludin molecule on the adjacent cell. The distal C-terminus forms a coiled-coil domain that participates in protein–protein interactions, binding directly to the intracellular protein ZO-1 [20]. VEGF promotes PKC-dependent serine/threonine phosphorylation of occludin [21, 22], which causes dissociation from ZO-1 [23],

disruption of the TJ and increased permeability. This effect may explain the activity of PKC inhibitors against vascular leakage in diabetic retinopathy (DR) [21].

The occludin content at the TJ correlates with the tightness of the barrier, with higher levels in cells known to have a tight barrier, such as arterial endothelial cells and brain endothelium [14, 24, 25]. Occludin expression appears 1 week postnatally (in rat models), which correlates with maturation of the barrier [5, 24]. Suppression of occludin expression (using antisense technology or siRNA) results in decreased barrier capacity to solutes [25, 26]. In rats with streptozotocininduced diabetes, decreased occludin content in the retina is noted and correlates with increased BRB permeability [14, 27, 28]. The localization of occludin also changes from continuous cell border to interrupted, punctate immunoreactivity in the arterioles [27]. This change in localization is associated with increased occludin phosphorylation at Ser490, which lies in the coiled-coil domain, and abolishes binding to ZO-1 [14]. In addition to VEGF [21], other stimuli that promote occludin phosphorylation and internalization are lysophosphatidic acid [29], histamine [29], oxidized phospholipids [30], and shear stress [31]. Conversely, hydrocortisone suppresses occludin phosphorylation, increases occludin expression and reduces BRB permeability [32], supporting the use of corticosteroids for the treatment of macular edema in DR [33].

Although occludin is an important component of TJs, it appears that it is not totally indispensable [34]. Occludin-deficient cells can still form functional TJs that recruit ZO-1, and occludin knock-out mice are viable, with TJs that appear morphologically normal and have normal transepithelial resistance (TER, a measure of permeability; although there is evidence of dysfunction of tissues that require barrier formation, such as testicular and gastric mucosa) [35]. Therefore, it appears that a high degree of redundancy in TJ composition exists, which can be explained by

the fact that the carboxy tail of claudins can interact with ZO-1, -2, and -3 and recruit them to the TJ [36], thus substituting to a major degree for the role of occludin.

Claudins

The claudin family comprises at least 24 members that are differentially expressed in various tissues. Claudins are 20to 27-kDa proteins with four transmembrane domains, two extracellular loops, and a short carboxy intercellular tail. Different cell types express different combinations of claudins [37]. The claudin expression pattern determines the barrier properties of individual TJ strands and is dynamically regulated during development, under normal conditions to respond to the selective permeability needs of the tissues, and during disease [37]. Claudins form both homopolymers and heteropolymers and bind across adjacent membranes, forming the TJ backbone [37–39]. Not all claudin combinations are compatible to form a functional TJ, and overexpression of an incompatible claudin type can result in a leaky TJ. For example, MDCK I cells normally express claudin-1 and claudin-4, and their TER values fall dramatically after overexpression of claudin-2, but not claudin-3 [40]. Several claudins participate in the formation of ion-selective channels, and genetic defects in these claudins are associated with disorders of ion transport and aberrant barrier function [37].

Claudin-5 is a critical component of TJs between endothelial cells, and its expression in the plasma membrane of retinal microvascular endothelial cells is significantly reduced under hypoxic conditions [41]. Inhibition of claudin-5 expression by RNAi resulted in a reduction of transendothelial electrical resistance, indicating a critical role of claudin-5 in the barrier property [41]. In claudin-5-deficient mice, the blood-brain barrier is selectively affected against small molecules (<800 Da), but not larger molecules [42]. In experimental autoimmune uveitis, expression

of TJ proteins claudin-1, -3 and occludin-1 in the retina was found to be decreased [43].

Intracellular Tight Junction Proteins

Zonula Occludens Proteins

The 225-kDa ZO-1 was the first TJ protein to be characterized [44], and the related ZO-2 and ZO-3 were subsequently identified [45–47]. ZO- 1, ZO-2 and ZO-3 are intracellular adaptor proteins that associate with the cytoplasmic surface of the TJ and are composed of three PDZ domains (PDZ1, PDZ2, PDZ3), one SH3 domain, and one guanylate kinase (GUK) domain that belong to the membrane-associated GUK protein family [14, 37, 46, 48]. ZO-1 is the central organizer of the TJ and forms a molecular scaffold that links the TJ to the cytoskeleton [5, 14]. Via its SH3-GUK region, it binds to occludin [49]. Via its PDZ-1 region, it binds to the COOHterminal tail of the claudins [36], and via its PDZ-2 domain it binds to ZO-2 [50]. ZO-1 and ZO-2 bind directly to actin filaments, thus linking the TJ to the cytoskeleton [47, 51]. Removal of all three ZO proteins revealed that at least one of ZO-1 or ZO-2 is necessary for TJ formation and establishment of barrier properties [14, 52]. The cell membrane expression of ZO-1 in various cells correlates with the barrier properties of these tissues. In endothelial cells from various tissues with potent barriers, VEGF induces tyrosine phosphorylation and redistribution of ZO-1 from the cell border to the cell interior [22, 53].

Pathophysiology of the Blood-Retinal Barrier

Breakdown

Two major pathways are responsible for hyperpermeability of retinal vessels in diseases of the retina, i.e. transcellular and paracellular transport [54]. Transcellular flux is very limited in BRB under normal conditions, due to the absence of fenestrations [5] and decreased pinocytosis.

However, this may change under pathologic conditions that lead to BRB breakdown. In a model of VEGF-induced retinopathy with microvascular leakage in monkeys, there was an increase in the number of pinocytotic vesicles at the endothelial luminal membrane, but no fenestrations or vesiculovacuolar organelles were found in the endothelial cells by electron microscopy [55]. The endothelium-specific antigen PAL-E, associated with transport vesicles in nonbarrier endothelium, which is almost absent from barrier capillaries in the normal brain and retina, is markedly and uniformly present in areas of vascular leakage in this model, as well as in human post-mor- tem eyes of individuals with DR [56–58].

Mechanismsofincreasedparacellularfluxthat lead to BRB breakdown are explained below.

Endothelial Cell Division or Cell Death

Large gaps in the capillary endothelial cell layer can be caused by endothelial cell division or cell death [5, 59], allowing leakage of fluid, protein and lipids. Endothelial cell death leads to the formation of acellular capillaries and pericyte ghosts. In a streptozotocin-induced model of diabetes, leukocyte adhesion to the diabetic retinal vasculature led to leukocyte-mediated, Fas- FasL-dependent apoptosis of endothelial cells [60]. Inhibition of FasL activity with a neutralizing antibody potently reduced retinal vascular endothelial cell injury, apoptosis, and BRB breakdown, but did not diminish leukocyte adhesion to the diabetic retinal vasculature [60]. Other causes of endothelial cell death could be oxidative stress [61–63], and advanced glycation endproducts (AGEs) [64]. BRB breakdown can also occur as a result of leukocyte extravasation during retinal inflammation [43].

Pericyte Loss

Pericytes are smooth muscle cells that form a sheath around the capillary endothelial cells. They help maintain vascular tone, provide structural support to the endothelium and release

growth factors necessary for endothelial cell survival [65]. Pericyte loss is a hallmark of early DR [66]. Pericytes exposed to high glucose concentrations exhibit increased expression of Bax and TNF-α, and undergo apoptosis [67–69]. Other mechanisms of pericyte loss include oxidative stress, toxicity from polyols, AGEs and heavily oxidized-glycated LDL, saturated free fatty acids, upregulation of protein kinase C, and focal leukostasis [65, 66, 70–77].

Tight Junction Disassembly

Disassembly of the TJs is frequently observed and leads to barrier breakdown in various pathologic conditions. This can be attributed to changes in TJ protein content and/or cellular localization. Vascular segments at the early stage of vascular formation and regression have decreased occludin expression [78]. Moreover, it has been known for the past 20 years that phosphorylation of TJ proteins modulates its permeability [79, 80].

Histamine causes a reversible concentrationdependent reduction of ZO-1 protein content in cultured retinal vascular endothelial cells [81, 82]. In the same model, high glucose (20 mm) and low insulin (10–12 m) reduced ZO-1 protein content, while astrocyte-conditioned medium increased ZO-1 protein content [82].

VEGF also has potent effects on TJ protein expression and localization. In a rat model of strep- tozotocin-induced diabetes, BRB permeability was increased and retinal occludin content decreased, an effect that was reproduced in bovine retinal endothelial cells treated in culture with VEGF [83]. In brain microvessel endothelial cells (BMECs), VEGF increased sucrose permeability and caused a loss of occludin and ZO-1 from the endothelial cell junctions and changed the staining pattern of the cell boundary. Western blot analysis of BMEC lysates revealed that the level of occludin but not of ZO-1 was lowered by VEGF treatment [84]. Phosphorylation of occludin reduces its ability to bind ZO-1, ZO-2, and ZO-3 [85]. VEGF promotes PKC-dependent