VASCULAR PERMEABILITY IN DIABETIC RETINOPATHY

The cause of visual loss in diabetic retinopathy remains unclear but likely involves loss of proper cellular interaction between the neural retina and retinal vasculature [83]. Changes in blood vessel permeability and macular edema consistently rank as the top clinical correlates associated with loss of vision [84, 85]. Indeed, central macular thickness, as measured by optical coherence tomography, and fluorescein leakage combined with age account for 33% of the variation in visual acuity [85]. Further, the location, severity, and duration of macular edema are all linked to visual loss [86]. Alterations to the BRB are believed to contribute to retinal macular edema with increased fluorescein permeability related to the progression of macular edema [87, 88]. Collectively, these clinical studies demonstrate a strong correlation with alterations to the BRB, increased macular edema, and loss of vision in patients with diabetes. It should also be noted, however, that other factors clearly contribute to vision loss in diabetes.

Vascular changes in diabetic retinopathy are due, at least in part, to elevated VEGF expression [89–94]. Indeed, recent clinical studies using anti-VEGF antibody therapy improved visual acuity in combination with laser compared to laser treatment alone [95]. In addition to VEGF, other cytokines likely also contribute to vascular changes in diabetic retinopathy. Increased levels of interleukin-1 beta (IL-1b (beta)) and tumor necrosis factor-alpha (TNF-a (alpha)) are increased in the vitreous of diabetic patients with proliferative diabetic retinopathy [96, 97] and in diabetic rat retinas [98–100], while leukostasis has been observed in response to elevated intracellular adhesion molecule-1 expression in diabetic rodents [101]. Furthermore, proteomic analysis of vitreous from patients with diabetic retinopathy reveals increased carbonic anhydrase I likely as a result of retinal hemorrhage and erythrocyte lysis [102]. The pH increase driven by carbonic anhydrase drives kallikrein activation leading to bradykinin production and permeability of the retinal vasculature as demonstrated by carbonic anhydrase I intravitreal injection. Therefore, multiple factors contribute to the increased retinal vascular permeability in diabetic retinopathy. Changes in both growth factors and inflammatory cytokines may induce alterations in the vascular barrier properties by distinct mechanisms over the course of diabetes. Thus, understanding the mechanisms of vascular permeability in diabetic retinopathy will allow the development of rationale therapies targeting specific disease characteristics or potentially identifying common mechanisms shared by the variety of cytokines altered in diabetic retinopathy.

VEGF-Induced Regulation of Endothelial Permeability

Both VEGF treatment of endothelial cells and induction of diabetes alter occludin content and localization associated with alterations in barrier properties. Studies on rats with streptozotocin-induced diabetes with 3-month duration reveal decreased occludin content and immunostaining at cell borders concomitant with increased BRB permeability. This change in occludin content can be recapitulated in bovine retinal endothelial cells (BREC) treated with VEGF [103]. Immunohistochemical analysis of occludin in diabetes or after addition of VEGF demonstrates that occludin localization at the cell border changes specifically at regions of paracellular permeability [54]. In this study, fluorescently labeled concanavalin A was perfused through control and diabetic or control and VEGF-treated retinas that were fixed to prevent active transport and preserve

protein localization. Concanavalin A does not bind endothelial cells directly but decorates regions where pores have formed that allow transport of the lectin to the endothelial basement membrane. Immunohistochemical analysis revealed that concanavalin A stained the basement membrane specifically at regions of low or absent occludin border staining, suggesting that redistribution of occludin away from the cell border created regions of paracellular permeability. Likewise, treatment of RPE cells with hepatocyte growth factor (HGF) reduced tight junctions, decreased TER, and increased diffusion of fluorescently labeled marker from the apical to basolateral membrane. After 6 h of HGF treatment, occludin, claudin-1, and a-catenin were redistributed from the membrane to the cytoplasm, and ZO-1 immunostaining was reduced [104]. Together, these studies demonstrate that changes in occludin are associated with altered permeability in the retina and suggest that occludin contributes to regulation of paracellular permeability in retinal endothelial cells.

Occludin Phosphorylation and Permeability

While gene deletion and knockdown of occudin expression reveal occludin is not necessary for formation of tight junctions, the observed changes in occludin content and localization associated with changes in barrier properties suggest occludin contributes to regulation of barrier properties. Recent studies suggest phosphorylation of occludin acts as a molecular switch to regulate endothelial barrier properties. Treatment of endothelial cells with VEGF [105, 106], cytokines [107], oxidized phospholipids [108], monocyte chemoattractant protein-1 (MCP-1 or CCL2) [109, 110], or shear stress [111] increased both serine/threonine phosphorylation of occludin and permeability. Furthermore, diabetes increases occludin phosphorylation in the rat retina similar to the VEGF-induced increase in BREC [106].

Phosphorylation of occludin leads to ubiquitination and subsequent endocytosis regulating endothelial barrier properties. The use of two-dimensional gel electrophoresis in BREC demonstrates that occludin is basally phosphorylated on two residues, and growth factor stimulation leads to phosphorylation at three additional sites [106]. Using mass spectrometry of occludin immunoprecipitated from vascular endothelial cells, Sundstrom et al. identified five putative occludin phosphosites and demonstrated at least one of these sites: Ser490 was VEGF responsive as shown by the use of a phosphospecific antibody [112]. This Ser490 phosphorylation allows subsequent ubiquitination of occludin by the E3 ligase Itch and endocytosis of the transmembrane protein by binding epsin, eps15, and Hrs, which possess ubiquitin interacting motifs and chaperon occludin through endocytosis [113]. Importantly, mutating Ser490 to alanine (S490A) prevented both occludin ubiquitination and VEGF-induced permeability, while expressing an occludin-ubiquitin chimeric protein creates leaky endothelial junctions. Thus, the carboxy-terminal tail of occludin can be phosphorylated and subsequently ubiquitinated, directing occludin into the endocytosis pathway and regulating endothelial barrier properties, potentially by controlling the localization of other junctional proteins such as the claudins.

While occludin phosphorylation and ubiquitination are necessary steps for VEGFinduced permeability, additional junction alterations are likely involved in the process. Recently, ubiquitination of claudins has also been observed in epithelial cells with the

E3 ubiquitin ligase LNX1p80 regulating claudin internalization and lysosomal degradation [114]. Further, in endothelial cells without tight junctions, the phosphorylation and endocytosis of VE-cadherin is an essential step to regulate barrier properties [115]. Additionally, the ubiquitin ligase Hakai ubiquitinates E-cadherin and induces endocytosis [116]. While the mechanisms controlling barrier properties are complex, posttranslational modifications regulating endocytosis of junctional components provide important mechanisms of permeability regulation.

Protein Kinase C in Regulation of Barrier Properties

Key mediators of BRB homeostasis and diabetes-induced vascular abnormalities include the Protein Kinase C (PKC) family [117]. Alterations of PKC isoforms during diabetes may result from hyperglycemia, de novo synthesis of diacylglycerol (DAG), advanced glycation end products (AGEs), increased expression of growth factor/inflammatory cytokines, and to a generally altered redox state [118]. As a member of the larger protein kinase AGC super family, PKC isozymes regulate essential signaling pathways in various tissues controlling proliferation, differentiation, survival, and cell growth (reviewed in [119–122]). There are three main classes of PKC isoforms based on their cofactor requirements. The classical PKC isoforms, a (alpha), bI, bII (betaI, betaII), and g (gamma), require Ca2+ and diacylglycerol (DAG) for activation. Novel PKC isoforms, d (delta), e (epsilon), h (eta), and q (theta), require DAG; while the atypical PKC isoforms, z (zeta), i (iota) and l (lamda), require neither DAG or Ca2+ to become activated [122].

Evidence for a role of PKC isoforms in vascular permeability and increased flux of macromolecules began in the late 1980s and early 1990s [123, 124]. Treatment of bovine pulmonary artery endothelial cells with phorbol 12-myristate 13-acetate (PMA), an activator of classical and novel PKC isoforms, leads to an approximately twofold increase in 125I-albumin permeability [123]. Additionally, PMA and diacylglycerol treatment of bovine aortic endothelial cells alters 14C-sucrose and 3H-inulin flux but not 125I-polyvinyl pyrrolidone (360 kDa) permeability, indicating PKC isoforms control paracellular permeability [124].

Diabetes-induced vascular permeability can be partly attributed to increased classical PKC activity. PKC activity is altered in the diabetic rat retina, BREC, and in bovine retinal pericytes (BRPs) [117]. Oral administration of LY333531, a specific PKCb (beta) inhibitor with low nanomolar potency similar to ruboxistaurin, ameliorates the diabe- tes-induced effect on retinal blood flow [125]. Membrane translocation and activation of PKCa (alpha), b (beta)II, and d (delta) isoforms in response to VEGF have been observed in vivo [126], and this translocation was blocked by oral administration of the PKCb (beta) inhibitor [127]. Mechanistically, increased activity of classical PKC isozymes leads to tight junction deregulation, cytoskeleton rearrangements, and endothelial permeability [106, 128]. Data from our laboratory demonstrates VEGF-induced occludin phosphorylation, and ubiquitination requires PKCb (beta) (manuscript in preparation). Furthermore, PKCa (alpha) mediates hyperglycemia-induced porcine aortic endothelial cell permeability demonstrated by RNAi knockdown [129]. Collectively, these data implicate classical PKC isoforms mediate vascular endothelial permeability induced by diabetes.

Although classical PKC isoforms contribute to VEGF-induced endothelial permeability, other signaling pathways also contribute to control of the BRB. Studies of primary retinal endothelial cell culture assays show an incomplete attenuation of VEGF-induced endothelial permeability via classical PKC inhibition [106]. In addition, tumor necrosis factor a (alpha) induces endothelial permeability over 6 h but is unaffected by classical PKC inhibitors (manuscript under review). Together, these data suggest concurrent or alternative signaling pathways may also contribute to the vascular permeability observed in diabetic retinopathy.

In addition to classical PKC isozymes, novel PKCs are implicated in mediating dia- betes-induced alterations of BRB homeostasis. PKCd (delta) translocates to the membrane fraction of retinal lysates of diabetic mice indicative of PKCd (delta) activation [130]. Geraldes et al. identified Src homology 2 domain-containing phosphatase-1 (SHP-1), a protein tyrosine phosphatase, as a downstream target of PKCd that leads to platelet-derived growth factor beta-receptor (PDGFb (beta) receptor) dephosphorylation. PDGFb (beta) is a survival signal for retinal pericytes allowing for activation of Akt, which is essential to pericyte survival [131]. Reduced PDGFb receptor signaling results in diabetes-induced pericyte apoptosis, which increases vascular permeability in the diabetic mouse retina [130]. In addition, PKCd mediates AGE-induced permeability in human retinal endothelial cells (HREC) as shown through the use of PKCd small molecule inhibitors and siRNA studies which prevent the AGE-induced alterations to ZO-1 and ZO-2 protein expression [132].

In addition to the well-established contributions of classical and novel PKC isoforms to diabetes-induced junctional deregulation and vascular permeability, a role for the atypical PKC (aPKC) isoforms is emerging. The aPKC isoforms act downstream of both the phosphatidylinositol 3-kinase (PI3-K) and the small Rho GTPases family members in response to growth factors, leading to proliferation, differentiation, and cell polarity/apical-basolateral orientation [73, 133]. Additionally, aPKC isoforms are critical for the establishment of primordial junction development and the regulation of junction complexes in both endothelial and epithelial cells [134, 135]. VEGF administration leads to a twofold increase in PI3-K activity as well as transiently activating small Rho GTPases such as Cdc42, Rac1, and Rho, contributing to endothelial permeability in endothelial cells [126, 136]. Therefore, aPKC isoforms may play a critical role in the regulation of growth-factor-induced vascular permeability. Data from our laboratory demonstrates overexpression of PKCz (zeta), an atypical PKC isoform, potentiates the effect of VEGF on permeability, whereas kinase dead-mediated competitive inhibition of PKCz (zeta) blocks VEGF-induced permeability in BREC. Importantly, aPKC inhibition prevents TNFa-induced endothelial permeability and prevents loss of tight junction proteins claudin-5 and ZO-1 and cell border disorganization (manuscript under review). Together, these studies demonstrate aPKC isoforms contribute to VEGF and TNFa-induced permeability, elucidating a common signaling mechanism in diabetic retinopathy. Collectively, these data show a contribution of classical, novel, and atypical PKC isoforms in the control of retinal vascular permeability (Fig. 3).