THE JUNCTIONAL COMPLEX

Formation of a well-developed tight junction creates the barrier across both the retinal vasculature and RPE as shown by electron microscopy studies. Horseradish peroxidase used as a stain in electron microscopy diffuses only to the tight junction in brain cortical capillaries, while in other tissues without tight junctions, this marker diffuses out of the vascular lumen [15]. Similar studies in the retina with dyes reveal that tight junctions mediate the BRB, preventing solute flux into the retinal parenchyma [16, 17]; however, the adherens junctions are essential to development of the barrier, and likely influence the formation of the tight junction [18–21]. Further, in blood-brain and BRBs, the tight junctions and adherens junctions are indistinguishable at the ultrastructural level [22, 23].

Tight junctions are composed of over 40 proteins encompassing transmembrane proteins, intracellular scaffolding proteins, and signaling proteins, acting in concert to influence barrier properties [24]. The transmembrane proteins include occludin, claudin family members, tricellulin, and the junctional adhesion molecules (JAMs). The transmembrane proteins are linked to the cytoskeleton via an interaction with the scaffolding protein family zonula occludens (ZO). Together, these proteins create a barrier to paracellular flux and contribute to the BRB.

ZO Proteins

The zonula occludens, or ZO, family members bind to both transmembrane structural proteins and regulatory proteins and organize the junctional complex. ZO-1 (210– 225 kDa) was the first tight junction protein identified, and subsequent studies using coimmunoprecipitation identified the other ZO family members, ZO-2 (180 kDa) and ZO-3 (130 kDa) [25–29]. ZO-1 and ZO-2 also associate with the adherens junctions [30] potentially as a first step in formation of tight junctions. ZO proteins are members of the membrane-associated guanylate kinase (MAGUK) family and are characterized by the presence of three PDZ domains, one SH3 domain, and a GUK domain [31]. ZO family members are also characterized by the presence of an acidic domain, a basic domain, a leucine zipper, and a proline-rich C-terminus [25, 27, 32, 33].

The contribution of ZO-1 in junctional protein organization has been demonstrated in cell culture and gene deletion studies. The calcium switch assay allows rapid disassembly of tight junctions followed by reassembly upon return of calcium to the medium. The use of siRNA to reduce ZO-1 expression results in reduced tight junction assembly in the calcium switch assay [34, 35]. ZO-2 also contributes to junction assembly and permeability as demonstrated by ZO-2 silencing which leads to a reduction in TER values in the calcium switch assay, without affecting mature TJs and increased permeability to 70 kDa dextran [36]. Deletion of ZO-1 and ZO-2 in a cell line lacking ZO-3 led to a complete loss of tight junction formation [37]. In vivo, ZO-1 [38] and ZO-2 [39] gene deletions have been described and both are lethal very early in mouse embryogenesis. However, distinct phenotypes suggest nonredundant function for these isoforms. ZO-1 gene deletion caused developmental defects in mouse embryo, yolk sac, and allantoic membrane vasculature, suggesting a role for ZO-1 in angiogenesis [38]. Interestingly, ZO-3 deletion does not impart lethality [39].

Claudins

The claudin family consists of 24 distinct proteins that form the tight junction seal between neighboring cells particularly regulating ion flux. The claudins are 20–27 kDa and possess four membrane-spanning domains with two extracellular loops and the N- and C-terminus in the cytoplasm [40–42]. The C-terminus of claudins is essential to both their stability and their membrane targeting [43, 44]. Of important note, all claudins possess a YV sequence as the final two amino acids that is necessary for their interaction with ZO [45, 46].

Claudin family expression patterns vary from tissue to tissue, and expression of different claudins confers specificity of barrier properties. In Madin Darby canine kidney (MDCK) cells, claudin-1 overexpression increases TER values by fourfold concurrent with a decrease in permeability to small and large molecules (4 and 40 kDa FITCdextrans) [47]. Claudins not only increase barrier properties but can also form chargeselective paracellular ion channels. For example, claudin-16 controls magnesium flux in the loop of Henle in the kidney and genetic defects of this claudin are associated with loss of magnesium [48]. Mutations of claudin-16 alter the sodium flux reducing magnesium transport potential [49]. A role for claudins in creating a charge specific barrier was demonstrated by mutational analysis. Exchanging two acidic residues in the first extracellular loop, Asp55 and Glu64, to create basic residues (D55R and E64K) in the extracellular loop of claudin-15 changes charge selectivity for paracellular permeability from sodium to chloride [50].Finally, siRNA studies altering expression of claudins-2, 4, and 7 can differentially alter cation or anion permeability [51]. Together, these studies demonstrate claudin expression, provide specific ionic barriers, and provide charge-selective paracellular channels.

A model for claudin barrier formation has recently been proposed based on transfection studies and is distinct from many of the schematics of tight junctions previously presented. Overexpression of claudin-5 in human embryonic kidney (HEK) cells, a cell type that typically does not express tight junctions, leads to formation of strands of tight junctions in the plasma membrane [52]. The investigators used mutational analysis to distinguish trans-interactions or interactions between claudins on adjacent cells as opposed to cis-interactions or interactions between claudins within a cell. By expressing fluorescent-tagged claudin-5 and performing a combination of live cell imaging, fluorescence resonance energy transfer, and scanning electron microscopy (SEM), the investigators were able to identify specific amino acids in the second extracellular loop responsible for trans-interactions. The position of these amino acids combined with SEM images led the investigators to propose a model in which 2 claudins first form a dimer within a membrane (cis-interaction). This dimer then interacts by loop 2 interactions to another dimer pair across the membrane (trans-interaction). This model of claudin interaction literally forms a zipper with the dimer pair of claudins interdigitating to create the tight junction seal (Fig. 1).

In the RPE, the expression of claudins-1, 2, and 5 have been detected in the developing chick embryo by embryonic day 14 [53]. Further, claudins-1, 5, and 15 are expressed in endothelial cells [45], and claudin-5 is expressed in the retinal vasculature [54]. Several studies have examined the effect of loss of claudins on barrier properties. Claudin-1 deletion in mice is lethal within 1 day postbirth as a result of excessive water loss through

Fig. 1. Proposed interactions between claudins at the tight junction. The claudin proteins form a zipper-like structure at the tight junction by alternating cis- and trans-interactions. Claudin proteins within the same cell form a cis-interaction forming a dimer pair. This dimer of claudins interacts with another dimer pair in adjacent cells through loop 2 interactions forming a trans-interaction between the two pair. Adopted from Piontek et al. [52].

the skin [55]. A claudin-5 knockout mouse was created, which developed normally and upon birth appeared grossly normal. However, the mice died within 10 h after birth due to increased permeability of small molecules (<800 Da) across the blood-brain barrier [56]. Finally, evidence for claudin-11 in the neural tissue is found from gene deletion studies. Mice deficient for claudin-11 showed severe neurological disorders and male sterility as a result of loss of tight junctions within the CNS and Sertoli cells [57]. Unfortunately, there is little information regarding specific changes to claudins in diabetic retinopathy. A study of mRNA content of claudins-1 and 5 reveal claudin-1 mRNA first increases at 6 weeks then decreases by 12 weeks postinduction of diabetes, while claudin 5 mRNA is decreased modestly at both time points [58]. Collectively, these studies indicate that claudins are essential for tight junction function, creating charge specific barriers while providing ion selective paracellular channels across the barrier. The regulation of claudin function in diabetic retinopathy remains an area for further research.

Junctional Adhesion Molecules

The junctional adhesion molecules, or JAMs, are glycosylated single-pass transmembrane proteins, with the C-terminus located intracellularly, and an extracellular N-terminus with two immunoglobulin (Ig)-like domains [42]. The JAMs are subdivided into two groups based on sequence homology [59, 60]. The first subgroup, composed of JAM-A, JAM-B, and JAM-C, directly interacts with ZO-1 and PAR3, a protein required for cell polarity, through a C-terminal class II PDZ-binding domain motif [61–63]. The second subgroup, which is composed of coxsackie and adenovirus receptor (CAR), JAM-4, and endothelial-cell-selective adhesion molecule (ESAM) contains a class I PDZ-binding

domain motif [42]. CAR and JAM-4 bind with the Ligand-of-Numb protein X1 by this PDZ-binding domain [64, 65], while JAM-4 and ESAM interact with the MAGUK protein [66, 67]. JAMs are also able to form homodimers and heterodimers through the extracellular domains. Specifically, JAM-A, JAM-B, and JAM-C interact with the integrins aLb2, a4b1, and aMb2, respectively [59, 60, 68].

JAM-A is necessary for junction resealing in both epithelial and endothelial cells. Specifically, studies demonstrate that monoclonal antibodies against JAM-A significantly inhibit junction recovery in a calcium switch assay as measured by transepithelial electrical resistance (TER) [69–71]. JAM-A also is involved in proper polarity maintenance [72] likely through its direct and specific interaction with PAR-3 [61, 62, 73]. Finally, ESAM is exclusively localized to endothelial cells [74], and its loss augments VEGF-induced permeability [75].

Occludin and Tricellulin

Occludin was the first transmembrane TJ protein discovered and is a 522-amino-acid protein of 55.9 kDa, and like the claudins, has four transmembrane domains [76]. However, the sequence and structure of occludin is sufficiently distinct from claudins, suggesting a unique role of this protein in tight junctions. A second gene product, tricellulin or MARVEL D2, has recently been identified as a 555-amino acid protein localized specifically at regions where three cells make contact [76]. Tricellulin and occludin share homology in the MARVEL domain across the tetra-transmembrane regions. MAL and related proteins for vesicle trafficking and membrane link or MARVEL domains are present in vesicle transport proteins such as MAL, which are essential for apical trafficking of membrane and secretory proteins in epithelia and also in the neural vesicle proteins synaptophysin and synaptogyrin [77]. In epithelial cells, occludin is enriched at bicellular junctions, while tricellulin is enriched at tricellular junctions; however, upon knockdown of occludin, tricellulin can be observed at bicellular junctions, which suggests occludin normally restricts tricellulin localization [78]. Little is known about tricellulin in endothelial cells, so this chapter’s primary focus will be on occludin.

Occludin content at the TJ correlates with barrier properties such that occludin is higher in cells with a tighter barrier, such as arterial endothelial cells and brain and retinal endothelium, and lower in cells known to have a more permeable barrier, such as venous endothelial cells and endothelial cells of nonneuronal tissues [79, 80] (Fig. 2). However, occludin does not provide a structural barrier for the tight junctions as do the claudins. Occludin knockout mice are viable and appear to form TJs but exhibit a number of abnormalities, including postnatal growth retardation, abnormalities in the testis leading to male infertility, and inability of females to suckle their young. Additionally, salivary gland abnormalities, thinning of compact bone, brain calcium deposits, chronic gastritis, and hyperplasia of the gastric epithelium are all a consequence of occludin gene deletion in mice [81, 82]. While a large number of additional studies including siRNA and overexpression studies suggest occludin contributes to barrier properties in a host of cell types (reviewed in ref. 2), the role of occludin in the barrier has remained difficult to understand. Recent studies suggest that occludin might not provide a direct structural component to the tight junction complex but rather act as a regulator of barrier properties.