Zonula Occludens (ZO) proteins

ZO proteins (ZO-1, ZO-2, and ZO-3) belong to the membrane-associated guanylate kinases (MAGUKs) superfamily and are characterized by their PDZ (PSD-95/Dlg/ZO-1), SH3 (Src-homology-3) and GuK (guanylate kinase) domains (Fig 3-10) [112]. ZO proteins are scaffolding proteins that recruit various molecules to TJs. They provide the structural basis for the assembly of TJ cytoplasmic plaques. Although the overall sequence similarity between three ZO proteins is low (Fig 3-11) [113], they share significant structural similarity:

three PDZ domains (PDZ1, -2, and -3), one SH3 domain, and one GuK domain that are separated by unique regions (U1–6). Importantly, while all three ZO proteins share redundant functions they are not functionally identical [114].

PDZ domains consist of 80-100 amino acid residues. They are protein-protein interaction modules found in most species [115-117] that either form dimers or bind to carboxyl-termini of proteins containing PDZ-binding motifs [118]. ZO proteins can form homodimers or heterodimers via their second PDZ domain (PDZ2) and exist as independent ZO-1/ZO-1, ZO-2/ZO-2, ZO-1/ZO-2 or ZO-1/ZO-3 complexes [68,119-123]. ZO protein amino-termini directly interact with TJ transmembrane proteins. For instance, all three ZO proteins bind to claudins at carboxy-termini via their first PDZ domain (PDZ1) [68] while ZO-1 binds to JAM via its third PDZ domain (PDZ3) [124]. Using ZO-1 domain-deletion transgenesto rescue the ZO-1/ZO-2 depletion phenotype in MDCK II cells, a recent study revealed that PDZ1, but not PDZ2, is essential for the normal organization of TJs and the AJ cytoskeleton. On the other hand, PDZ2 is important for normal TJ permeability [125].

Figure 3-10 Structural domains and protein-binding sites of ZO-1, ZO-2 and ZO-3. ZO proteins interact with numerous molecules. Colored boxes highlight the binding sites of TJ-specific proteins (claudins, occludin and ZONAB) discussed in this thesis. (Modified from Lorenza Gonzalez-Mariscal et al. [112] )

Figure 3-11 Sequence similarities between full-length mouse ZO proteins and their respective domains. (Modified from Makoto Adachi et al. [113])

The GuK domain of MAGUK proteins lost its enzymatic characteristics over evolution due to the absence of amino acids responsible for GMP and ATP binding. Instead, it associates with SH3 domains and mediates protein–protein interactions [114]. For instance, the SH3-GUK intramolecular fold is conserved in ZO-1 and is required for normal TJ assembly kinetics [126,127]. SH3 domains, consisting of approximately 60 amino acid residues, are found in proteins involved in signal transduction, such as the Src kinase, and proteins that regulate membrane-cytoskeleton interactions [128]. Signaling proteins, such as the serine protein kinase ZAK (ZO-1 associated kinase-1) and the transcription factor ZONAB (ZO-1-associated nucleic acid binding proteins) bind to the SH3 domain of ZO-1 [129,130].

More importantly, the SH3-U5-GuK-U6 region of ZO-1 functions as an integrated module to direct appropriate localization of other TJ proteins [127]. The U5 region of ZO-1 binds to occludin and targets ZO-1 to TJs, while U6 inhibits occludin binding to ZO-1. ZO-1 and ZO-2 directly bind to actin filaments via their C-termini [131,132], while ZO-3 is thought to

interact with F-actin via its N-terminus [133]. More than 400 proteins that potentially interact with ZO-1 were recently identified by biotinylation experiments using biotin ligase fused to either the N- or C-terminus of ZO-1 as bait [134]. That study indicated that ZO protein termini are associated with different but overlapping proteins. The N-terminus interacts more frequently with TJ transmembrane proteins and signaling proteins, while the C-terminus interacts with cytoskeletal proteins. This result confirms the hypothesis that ZO proteins are scaffolding proteins that link TJs to the actin cytoskeleton.

Zonula Occludens-1

ZO-1, approximately 220 kD in size, was the first identified TJ-associated molecule. This was achieved in 1986 thanks to a monoclonal antibody generated against a detergent-insoluble TJ-enriched membrane fraction isolated from mouse liver [135]. ZO-1 is resistant to nonionic detergents but can be solubilized from isolated membrane fractions either by 6 M urea or under conditions of high pH, suggesting that ZO-1 is a peripherally-associated membrane protein [136]. ZO-1 is present in both epithelial and non-epithelial cells [137]. It is specifically enriched at TJs of well polarized epithelial cells since it directly binds to TJ transmembrane proteins, such as claudins [68], occludin [138,139], JAM [124] and tricellulin [84]. In non-epithelial cells lacking TJs, such as fibroblasts and cardiac muscle cells, ZO-1 associates with AJs by directly interacting with -catenin [140] or by associating with gap junctions (GJs) via connexins [141,142].

The functional roles of ZO proteins were revealed by a series of deletion and down-regulation studies. Complete depletion of endogenous ZO-1 expression in Eph4 cells

had no effect on either TJ or AJ formation. However, TJ formation was delayed following Ca²⁺ depletion [143]. Similarly, shRNA- or siRNA-induced depletion of ZO-1 in MDCK cells delayed TJ formation [126] and slightly increased large solute permeability [144]. On the other hand, depletion of both ZO-1 and ZO-2 completely abrogated TJ assembly without affecting cell polarity, disrupted paracellular barrier function and increased expansion of the apical actomyosin contractile array located at the apical junction complex (AJC) [145,146].

Abrogated TJ formation can be rescued in ZO-1(KO)/ZO-2(KD) cells expressing either exogenous ZO-1 or ZO-2, suggesting that the roles of ZO-1 and ZO-2 in junction formation might be functionally redundant, at least to some extent [145]. These in vitro data indicate that ZO-1 is required for normal TJ formation and function.

The role of ZO-1 is not limited to TJ formation. During epithelial cell polarization, ZO-1 likely interacts with AJ proteins, such as α-catenin and AF-6/afadin, during AJ assembly prior to TJ protein recruitment [147-149]. ZO-1 is thought to be required for coupling AJ and TJ assembly.

ZO-1 knockout mice are embryonic lethal, owing to massive apoptosis in the notochord, neural tube area, and allantois at embryonic day (E) 9.5 and no viable embryos were found beyond E11.5 [150]. In wild-type embryos, ZO-1 and ZO-2 have very similar expression patterns, except in the yolk sac extraembryonic mesoderm that expresses ZO-1 but not ZO-2 or ZO-3. That study found that ZO-1 deficiency did not affect ZO-2/3 expression but did lead to mislocalization of endothelial JAM-A in the yolk sac, which might explain angiogenesis deficiency. These observations indicate that ZO-1 is essential to embryonic development in both the embryonic and extraembryonic regions.

In addition to its role in junctional complex formation and function, ZO-1 has also been shown to control gene expression by regulating intracellular ZONAB translocation [151].

ZONAB is a transcription factor that shuttles between TJs and nuclei [130]. Nuclear ZONAB promotes cell proliferation by regulating the expression of CycD1 and proliferating cell nuclear antigen (PCNA) [152]. In confluent MDCK cells, ZONAB binds to the SH3 domain of ZO-1, forming a complex with ZO-1 and CDK4 (cyclin-dependent kinase 4) at TJs [153].

ZONAB cytoplasmic sequestration inhibits G1/S phase transition by reducing CycD1 and PCNA expression levels. ZONAB has also been shown to repress proximal tubule cell differentiation by controlling megalin and cubilin expression [154], endocytic receptors that mediate protein reabsorption by the proximal tubule [155]. These studies indicate that ZO-1 might indirectly regulate cell proliferation and differentiation via ZONAB, at least in some cell types. Interestingly, a newly released study shows that depletion of ZO-1 in MDCK cells has no effect on ZONAB localization and activity, and that ZO-1-deficient Eph4 cells still show junctional ZONAB, indicating that ZO-1 may not be required for sequestration of ZONAB at junctions [156]. In addition, this study shows that three ZO proteins redundantly control ZONAB localization and stability [156].

Immunostaining revealed that ZO-1 is expressed in the nucleus of subconfluent, but not confluent, MDCK and LLC-PK1 cells [157]. It was also detected by Western-blot in the nucleus of microdissected rabbit renal tubules [158]. Detailed sequence analysis revealed that ZO-1 contains two nuclear localization signals (NLS) located in the PDZ1 and GuK domains [159]. However, its nuclear function is unclear.

Zonula Occludens-2

ZO-2, approximately 160 kDa in size, was identified in 1991 as a ZO-1-binding protein by immunoprecipitation in epithelial cells [160]. ZO-2 has a domain structure similar to that of ZO-1 and is concentrated at TJs of epithelial and endothelial cells and at AJs of nonepithelial cells, such as fibroblasts and cardiac muscle cells [119,161]. ZO-2 also associates with GJs by directly binding to connexins [162-164]. Its binding characteristics are shown in Figure 3-10 [112].

The roles of ZO-2 in regulating the assembly and function of TJs have been investigated in cultured cells by ZO-2 silencing. Transient and stable knockdown of ZO-2 in MDCK cells yield distinct phenotypes. Transient knockdown of ZO-2 by siRNA increased paracellular permeability, decreased TER during TJ formation and induced E-cadherin mislocalization [165]. However, ZO-2 knockdown using a tet-off system had no observable effect [144].

Similarly, downregulation of ZO-2 in Eph4 cells stably expressing shRNA had no discernible effect [145]. These conflicting results might be due to differences between transient and stable transfection systems.

Similar to ZO-1, ZO-2-deficient mice are embryonic lethal. Embryos lacking ZO-2 exhibit decreased proliferation at embryonic day 6.5 (E6.5) and increased apoptosis at E7.5, compared to wild-type embryos, and die shortly after implantation due to arrested growth during early gastrulation [166]. ZO-2-deficent mice die prior to ZO-1-deficent mice (at E10.5–11.5), suggesting that ZO-1 and ZO-2 do not play redundant roles during early embryo development. Microinjection of mouse zygotes with ZO-1 siRNA resulted in a more severe inhibition of blastococel formation than that induced by ZO-2 knockdown, indicating that ZO

proteins play different roles in blastocyst morphogenesis [167]. In addition, male ZO-2 chimeric mice displayed reduced fertility and defects in blood–testis barrier integrity in the absence of altered expression of other TJ proteins, indicating that ZO-2 plays an essential role in the development of extraembryonic tissues and that its role in embryonic development is non-redundant as compared to that of ZO-1 and ZO-3 proteins [168].

ZO-2 contains several nuclear localization and exportation signals (NLS and NES).

Mutation of any of these sites in canine ZO-2 is sufficient to induce nuclear accumulation or to impair nuclear import of ZO-2 protein [169,170]. ZO-2 is present in the nucleus and TJs of low-density MDCK cells and concentrates at TJs of confluent cultures [171]. Immunostaining also revealed nuclear ZO-2 expression in highly migratory endothelial cells [172]. Cell stress, such as heat shock and mechanical injury, increased ZO-2 nuclear localization in confluent epithelial and endothelial cells, suggesting that ZO-2 might regulate intracellular signaling pathways under these conditions [171,172].

A number of nuclear proteins have been found to associate with ZO-2, indicating that this protein might play a role in regulating gene expression. For instance, nuclear ZO-2 directly interacts with the nuclear matrix protein SAF-B (scaffold attachment factor-B) [172]

and associates with several transcription factors, such as Fos, Jun and C/EBP (CCAAT/enhancer binding protein) [173]. However, the role of ZO-2 in regulating gene expression via these nuclear proteins remains unclear. Further evidence shows that nuclear ZO-2 inhibits CycD1 transcription by interacting with the transcription factor c-Myc, which impairs binding of c-Myc with the enhancer box element contained in the CycD1 promoter [174]. ZO-2 overexpression in MDCK cells inhibits CycD1 protein synthesis and blocks cell

cycle progression at the G0/G1 phase [175]. Thus, ZO-2 directly regulates cell growth and proliferation.

Recently, ZO-2 has been shown to form a complex with YAP (Yes-associated protein) and its paralogue TAZ (transcriptional co-activator with PDZ domain) [176,177], two transcriptional co-activators of the Hippo pathway that controls organ size by coordinating cell proliferation and apoptosis [178,179]. Non-phosphorylated nuclear YAP and TAZ interact with a number of transcription factors, which promotes expression of proliferation-associated genes. Activation of the Hippo pathway leads to YAP and TAZ phosphorylation, which inhibits their nuclear translocation [178]. YAP interacts with the PDZ1 domain of ZO-2 via its own PDZ-binding motif, which facilitates YAP/TAZ shuttling between junctional complexes and the nucleus [176,180]. While ZO-2 overexpression in MDCK cells induces nuclear accumulation of ZO-2 and YAP, it attenuates YAP-mediated proliferative activity, presumably since the anti-proliferative effects of nuclear ZO-2 dominate the pro-proliferative effects of YAP [176]. In HEK-293 cells, ZO-2 cooperates with YAP to promote cell detachment and apoptosis [176,181]. TAZ was also shown to associate with the PDZ1 domain of ZO-1 and ZO-2 via its C-terminal PDZ binding motif, and ZO-2 overexpression inhibits TAZ-mediated transactivation [177]. These results indicate that ZO-2 might be a negative regulator of YAP and TAZ.

Zonula Occludens-3

ZO-3, approximately 130 kDa in size, was identified in 1993 as a protein that co-immunoprecipitates with ZO-1 in extracts of cultured MDCK cells [182]. Unlike ZO-1 and

ZO-2, which are expressed in both epithelial and non-epithelial cells, ZO-3 expression is restricted to epithelial cells in mouse and its expression is especially high in lung, liver and kidney [183]. ZO-3 is concentrated at TJs but not at cadherin-based AJs.

In MDCK cells, overexpression of the amino-terminal half of ZO-3 containing the three PDZ domains delayed assembly of both TJs and AJs during epithelial polarization [133], decreased the number of stress fibers and focal adhesions and increased cell migration [184].

However, TJ structure and function were unaffected in mouse ZO-3-deficient teratocarcinoma F9 cells [113]. The C-terminal region of ZO-3 can interact directly with p120 catenin [184] , a nucleo-cytoplasmic shuttling protein that regulates AJ stability and modulates the activities of several Rho GTPases [185]. ZO-3 may indirectly regulate Rho signaling by binding to p120 catenin [184].

Surprisingly, ZO-3 knockout mice are viable and fertile, and lack an obvious phenotype under normal laboratory conditions, suggesting that ZO-3 might be dispensable for mouse development [113,166]. As mentioned above, ZO-1, ZO-2 and ZO-3 may share functionally redundant properties. However, the true function of ZO-3 protein might be underestimated since ZO-1 and/or ZO-2 could functionally compensate for the lack of ZO-3. Interestingly, silencing of ZO-3 in zebrafish embryos revealed that ZO-3 is critical for epidermal barrier function and osmoregulation in early stages of zebrafish development [186].

ZO-3 is predicated to contain at least one conserved NES and several NLSs [170].

However, in contrast to ZO-1 and ZO-2, ZO-3 has not been observed in the nucleus so far.

Thus, its potential ability to shuttle between the cytoplasm and the nucleus is unclear.