Tight junctions at a glance

TJs contain two principal types of transmembrane components – tetraspan and single-span transmembrane proteins. The tetraspan proteins are occludin, tricellulin and the claudins, all of which are associated with the junctional intramembrane strands described above and are thought to have the same membrane topology (both the N- and C-termini in the cytosol). Tetraspan proteins form the paracellular permeability barrier and determine the capacity and the selectivity of the paracellular diffusion pathway (see below).

The claudins, which constitute a family of at least 24 members, are expressed in a tissue-specific manner. They are thought to be the main structural components of intramembrane strands (Furuse and Tsukita, 2006). The claudin composition of a junction determines the ion selectivity of the paracellular pathway, because changes in claudin expression correlate with alterations in conductivity for specific ions. Via their C-terminal cytoplasmic domains, claudins interact with several PSD95–DlgA–ZO-1 homology (PDZ)-domain-containing cytosolic proteins and these interactions are thought to be important for junction assembly. The extracellular loops of claudins have been proposed to create charge-selective paracellular aqueous pores that permit the passive diffusion of ions between cells (Krause et al., 2008; Van Itallie and Anderson, 2006); however, structural evidence for such pores is still missing and it is not clear how the extracellular domains of different claudin molecules cooperate to assemble continuous, ion-conductive pores.

Similar to the claudins, occludin is associated with intramembrane strands, although it is not required for their assembly. It regulates the paracellular diffusion of small hydrophilic molecules, has been linked to the formation of the intramembrane diffusion barrier and regulates the transepithelial migration of neutrophils (Aijaz et al., 2006). Either one of its N- and C-terminal cytosolic domains is sufficient to anchor the protein in the junction and to ensure a continuous junctional distribution. Although both domains interact with components of the cytosolic plaque, they have different functional properties – the N-terminal domain regulates neutrophil transmigration and the C-terminal domain regulates the paracellular diffusion of small hydrophilic tracers as well as targeting of occludin to the basolateral membrane. Similarly, the two extracellular loops of occludin regulate its accumulation at junctions as well as paracellular permeability. Occludin has also been linked to the regulation of various subcellular signalling pathways, such as MAP-kinase-dependent pathways and RhoA signalling. In the case of the regulation of Raf1 and TGFβ, this signalling function involves the second (more C-terminal) extracellular domain (Barrios-Rodiles et al., 2005; Wang et al., 2005; Yu et al., 2005). However, the underlying molecular mechanisms of many of the processes to which occludin has been linked remain to be determined and the functional relevance of most of its biochemical interactions is still unknown.

Tricellulin was first identified as a protein that is downregulated by the zinc-finger transcription factor Snail (Ikenouchi et al., 2005). In contrast to occludin and the claudins, tricellulin is enriched at the junctions between three epithelial cells. The suppression of tricellulin expression by RNA interference impairs the barrier function of TJs, which suggests that it is required for junction formation (Ikenouchi et al., 2005). In its C-terminal cytoplasmic domain, tricellulin shares a conserved domain with occludin that mediates binding to the tight junction protein 1 (TJP1, hereafter referred to as ZO-1), a cytosolic component of TJs (see below) (Riazuddin et al., 2006). Although this domain is affected in mutations that cause deafness, the role of the interaction of tricellulin with ZO-1 in the formation of junctions is unknown.

Most of the single-span transmembrane proteins that are associated with TJs, including the junctional adhesion molecules (JAMs), are members of the CTX/V_H-C₂ family of adhesion proteins. These proteins contain two immunoglobulin folds in their extracellular domain, one of the V_H- and one of the C₂-type (Bazzoni, 2003). Four JAMs – JAM-A, -B, -C and -D – have been identified, and all four can mediate homotypic cell-cell adhesion. Several of the JAM proteins, as well as the related proteins coxsackievirus and adenovirus receptor (CAR), CAR-like membrane protein (CLMP) and the endothelial-cell-selective adhesion molecule (ESAM), associate with TJs and interact with adaptor proteins of the cytoplasmic plaque (Bazzoni et al., 2000; Ebnet et al., 2003; Ebnet et al., 2000; Raschperger et al., 2004; Wegmann et al., 2004). JAMs regulate adhesion between leukocytes and endothelial cells, as well as the paracellular transmigration of leukocytes across the endothelium, and have been shown to regulate the development of cell polarity by binding to the evolutionarily conserved complex between PAR3, PAR6 and atypical PKC (aPKC) (Bradfield et al., 2007; Ebnet et al., 2004; Weber et al., 2007). The role of JAMs in cell polarisation is also important during spermatogenesis, because JAM-C-deficient spermatids fail to polarise and differentiate.

Another type of single-span membrane protein that is associated with TJs is Crumbs homologue 3 (CRB3), one of the three vertebrate homologues of the Drosophila melanogaster protein Crumbs (Lemmers et al., 2004; Makarova et al., 2003; Roh et al., 2003). CRB3 is present across the entire apical membrane and only a minor fraction of the entire cellular pool is associated with TJs. The cytoplasmic domain of CRB3 interacts with cytosolic adaptor proteins such as Pals1, which link it to the cellular machinery that regulates epithelial polarisation (Hurd et al., 2003; Lemmers et al., 2004). Consequently, the overexpression of CRB3 leads to delayed TJ formation in monolayers of Madin-Darby canine kidney (MDCK) cells and to disruption of epithelial morphogenesis. CRB3 also localises to the primary cilium and regulates ciliary morphogenesis (Fan et al., 2007).

Finally, Bves (a member of the Popeye-domain-containing family of transmembrane proteins), also localises to TJs in polarised epithelia and associates with ZO-1-based complexes (see below) (Osler et al., 2005). The downregulation of Bves in cultured cells reduces transepithelial resistance and junctional protein localisation at the membrane, which suggests that Bves contributes to the establishment and/or maintenance of epithelial integrity (Osler et al., 2005); however, the role of Bves in TJs is poorly understood and little is known about its interaction partners.

Intercellular junctions have similar building principles in different cell types and often share components. For example, many of the components that are associated with epithelial junctions are also found at neuronal synapses (Yamada and Nelson, 2007). These components often have similar functions, such as the regulation of the cytoskeleton to guide the assembly and disassembly of the junction or synapse. A major structural feature of junctions is that they contain a cytoplasmic plaque, which forms an interface between the junctional membrane and the cytoskeleton. In TJs, the cytoplasmic plaque functions in the regulation of adhesion and paracellular permeability, as well as in the transmission of signals from the junction to the cell interior to regulate cellular processes such as migration and gene expression. The cytoplasmic plaque is formed by a complex network of adaptors, scaffolding and cytoskeletal proteins that crosslink junctional membrane proteins and connect TJs to the actin cytoskeleton (see poster). The TJ plaque also recruits an array of signalling components that includes classical signalling proteins, such as kinases and phosphatases, as well as dual-localisation proteins that can reside at junctions as well as in the nucleus and provide a mechanism by which TJs can regulate gene expression (Guillemot et al., 2008; Matter and Balda, 2007; Paris et al., 2008).

ZO-1, the first TJ protein to be identified, is one of the most-studied TJ plaque components. It has the typical functional properties and domain structure of a scaffolding protein – it contains several protein-protein interaction domains, including three PDZ domains, one SH3 domain and one guanylate kinase homology (GUK) domain), as well as an F-actin-binding domain. ZO-1 interacts with claudins through its first PDZ domain, with ZO-2 or ZO-3 (two other TJ adaptors of the same protein family as ZO-1) through its second PDZ domain, with occludin through the GUK domain, and with actin and α-catenin through its large C-terminal domain (Fanning, 2001; Umeda et al., 2006). In addition, ZO-1 binds through its SH3 domain to several signalling proteins such as a serine/threonine protein kinase, the Y-box transcription factor ZONAB and the heat-shock protein Apg2, as well as Gα12, an α subunit of heterotrimeric G proteins (Matter and Balda, 2007; Meyer et al., 2002). These interaction partners can regulate each other; for example, Apg2 regulates the interaction between ZO-1 and ZONAB during the heat-shock response, which results in the nuclear accumulation and activation of ZONAB.

Other junctional proteins are present as components of evolutionarily conserved signalling complexes, which are based on the adaptors PAR3 and PAR6, or Pals1 and PATJ (Assemat et al., 2007; Wang and Margolis, 2007). Both of these complexes are found in both vertebrates and invertebrates and are crucial for the development of cell polarity and for epithelial morphogenesis. Both complexes bind to TJ membrane proteins: the PAR3-PAR6 complex binds to JAMs and the Pals1-PATJ complex binds to CRB3. The complexes are not independent but, instead interact with each other. For example, PAR6 also binds to Pals1 and CRB3 (Hurd et al., 2003; Lemmers et al., 2004), and ZO-3 – an interaction partner of ZO-1 – binds to PATJ (Michel et al., 2005; Roh et al., 2002). However, in most cases in which such interactions between components of different complexes have been identified, it has not yet been determined how common such interactions are and how they contribute to junction assembly and function.

The classical functions of TJs are the regulation of paracellular permeability and the formation of an apical-basolateral intramembrane diffusion barrier that helps to maintain cell-surface polarity. More recently, TJs have been linked to various signalling mechanisms that guide gene expression, proliferation and differentiation. TJ components also form complexes with the cellular machinery that regulates basolateral cell-surface transport (the Sec6-Sec8 complex); however, this does not appear to be an exclusive property of TJ proteins because these complexes also contain adherens-junction components, such as E-cadherin and nectin 2 (Yeaman et al., 2004). The actual structure that forms the intramembrane diffusion barrier is poorly understood; hence, we limit our discussion to the role of TJs in the regulation of paracellular permeability, as well as in signalling during epithelial proliferation and differentiation.

TJs allow the passive selective diffusion of ions and small hydrophilic molecules through the paracellular pathway across epithelia and endothelia. The molecular mechanisms that are responsible for selective ion permeability and for diffusion of small hydrophilic molecules are distinct, as many manipulations and regulatory mechanisms specifically affect only one of the two processes, or downregulate one while activating the other (Aijaz et al., 2006). For example, the overexpression of occludin or certain occludin mutants in cultured epithelial cells stimulates the paracellular diffusion of small hydrophilic molecules but increases transepithelial electrical resistance (a measure of transepithelial ion permeability) (Aijaz et al., 2006).

Occludin, tricellulin and the claudins are the main TJ membrane components that are involved in paracellular permeability. On the basis of observations in human disease, mouse models and cultured cell lines, it has been suggested that the claudin composition of TJs is a major determinant of the permeability properties of a tissue (Furuse and Tsukita, 2006; Van Itallie and Anderson, 2006). Absence of specific claudins can cause organ-specific defects, such as neurological, reproductive and renal defects. For example, mice that lack claudin 1 die after birth because of water loss across the skin, and the absence of claudin 5 causes leakage of small tracers across the brain endothelium (Furuse and Tsukita, 2006). Experiments with epithelial cell lines further suggest that different claudins favour paracellular diffusion of specific ions and, as mentioned above, these have lead to a model in which claudins form homo- and hetero-oligomers that engage in intercellular interactions to form paracellular aqueous pores. The ion-selectivity of these pores is determined by their claudin composition (Krause et al., 2008; Van Itallie and Anderson, 2006).

The process that mediates the paracellular diffusion of small hydrophilic molecules is less well understood and the actual mechanism by which such molecules permeate the junction is not known. However, the process is regulated by RhoA signalling and seems to require actinomyosin-driven processes (Aijaz et al., 2006; McKenzie and Ridley, 2007; Nusrat et al., 2000; Utech et al., 2006). It is therefore possible that dynamic rearrangements of intramembrane strands lead to paracellular diffusion. Occludin has been linked to the regulation of paracellular permeability of small hydrophilic molecules across cultured epithelial monolayers (Balda and Matter, 2000; Schneeberger and Lynch, 2004); it is thought that this regulatory mechanism involves phosphorylation events as well as the actin cytoskeleton because the C-terminal domain of occludin binds to protein kinases and lipid kinases, as well as to actin filaments and cytoskeletal linkers (Aijaz et al., 2006; Schneeberger and Lynch, 2004). Strikingly, live-cell-imaging experiments using GFP-tagged occludin have recently suggested that occludin diffuses within the junction, suggesting that occludin dynamics might contribute to paracellular diffusion (Shen et al., 2008). However, because the experiment involved an N-terminal GFP-tag and blocking the N-terminus is known to interfere with anchoring of occludin within the junction, it is not clear whether the observed dynamic properties indeed reflect physiological occludin behaviour (Huber et al., 2000).

The regulation of cell proliferation and polarisation is crucial for the development of differentiated tissues. Several studies have linked TJs to the regulation of cell proliferation and cell polarity. Similar to adherens junctions, TJs function in the suppression of proliferation (Gonzalez-Mariscal et al., 2007; Matter and Balda, 2007). Occludin suppresses oncogenic Raf-1 signalling (Wang et al., 2005) and ZO-1 interacts with ZONAB, thereby regulating gene expression, cell proliferation and epithelial morphogenesis (Matter and Balda, 2007; Sourisseau et al., 2006). ZO-2 localises to the nucleus and interacts with the DNA-binding protein scaffold attachment factor B (SAFB) as well as with several transcription factors (Gonzalez-Mariscal et al., 2007; Huerta et al., 2007; Traweger et al., 2003). TJs have also been linked to the regulation of RhoA-dependent proliferation through the junction-associated guanine-nucleotide exchange factor GEF-H1 and the junctional scaffolding protein cingulin, which binds to and inhibits GEF-H1 (Aijaz et al., 2005; Guillemot and Citi, 2006). It remains to be determined, however, whether and how these different TJ-associated signalling mechanisms are connected with each other and by which transmembrane proteins they are regulated. Strikingly, deficiencies in ZO-1 or ZO-2 expression are embryonic lethal in mice, which suggests that the two TJ proteins are important for development (Katsuno et al., 2008; Xu et al., 2008). Whether any of the signalling mechanisms that have been identified in cell culture experiments contribute to these phenotypes in vivo, however, is not clear.

Two evolutionarily conserved cell-polarity signalling pathways reside at TJs. Similar to D. melanogaster, the CRB3-Pals1-PATJ pathway regulates junction assembly and biogenesis of the apical membrane in vertebrate epithelial cells (Shin et al., 2006). The signalling pathway downstream of the complex has not yet been elucidated. The second conserved signalling module is the PAR3-PAR6-aPKC complex, which was originally described as a regulator of cytoplasmic partitioning in the early embryo of Caenorhabditis elegans. PAR3 associates with the cytoplasmic domain of JAMs, which results in the recruitment of the PAR3-PAR6-aPKC complex to cell-cell junctions (Bradfield et al., 2007; Ebnet et al., 2004; Weber et al., 2007). The complex functions as an effector of Cdc42, a Rho-family GTPase that is essential for epithelial-cell polarity and becomes activated during junction formation – binding of the complex to GTP-bound Cdc42 stimulates activation of aPKC and, consequently, the formation of mature TJs (Ebnet et al., 2004; Shin et al., 2006). In agreement with these observations, Cdc42 is crucial for the establishment of epithelial polarity during early mammalian development, and the aPKC isoforms PAR3 and PAR6 are necessary for the formation of the epidermal barrier (Helfrich et al., 2007; Wu et al., 2007). Interestingly, PAR3 also suppresses the activation of Rac1, another member of the RhoGTPase family, during junction formation by binding the guanine nucleotide exchange factor TIAM1 (Chen and Macara, 2005). This function, however, does not require PAR6 or aPKC, which suggests that PAR3 can act independently of PAR6 and aPKC.

TJ transmembrane proteins are affected in several inherited diseases, which suggests that the selectivity of the junctional diffusion barrier is physiologically important. For example, mutations in claudin 16 (which was originally called paracellin-1) and claudin 19 cause hypomagnesaemia (renal magnesium wasting) owing to a deficiency in paracellular magnesium resorption in the kidney (Konrad et al., 2006; Simon et al., 1999). The two proteins interact and are thought to form a paracellular cation pore. Similarly, mutations in claudin 14 and tricellulin cause hereditary deafness, which is likely to be the result of alterations in paracellular permeability (Riazuddin et al., 2006; Wilcox et al., 2001).

The WNK (with-no-K[Lys]) kinases WNK4 and WNK1, mutations in which cause hypertension (pseudohypo-aldosteronism type II) because of their effects on renal salt reabsorption and K⁺ excretion (Wilson et al., 2001), are also thought to act through claudins. Disease is caused by WNK4 alleles that have gain-of-function mutations and therefore hyperstimulate claudin phosphorylation, which results in increased paracellular Cl^– permeability and, subsequently, hypertension (Kahle et al., 2004; Yamauchi et al., 2004; Richardson and Alessi, 2008).

The expression of several TJ components is affected in various carcinomas. For example, the expression levels of ZO-1 and ZO-2 are dysregulated in different types of cancers and, in the case of breast cancer, low expression of ZO-1 has been correlated with a poor prognosis (Chlenski et al., 2000; Chlenski et al., 1999; Hoover et al., 1998; Kleeff et al., 2001; Martin et al., 2004; Morita et al., 2004; Resnick et al., 2005; Takai et al., 2005). Similarly, several junctional scaffolding proteins are bound and inactivated by viral oncogenes (Glaunsinger et al., 2001; Latorre et al., 2005). By contrast, ZONAB and its activating protein Apg2 are both upregulated in hepatocellular carcinomas, which suggests that this proliferation-promoting pathway is stimulated (Arakawa et al., 2004; Gotoh et al., 2004; Hayashi et al., 2002).

To what extent these alterations are a cause or a consequence of carcinogenesis is generally not clear. Nevertheless, claudin 1 has been shown to promote transformation and metastatic behaviour in colon cancer (Dhawan et al., 2005). The underlying molecular mechanism by which claudin 1 regulates migration is not clear. However, it might involve the association of claudin 1 with integrin-based complexes, similar to the role of claudin 11 in cell migration (Tiwari-Woodruff et al., 2001).

TJ proteins are targeted by several types of pathogens, and these interactions often lead to junctional dissociation and the loss of epithelial barrier function. For example, proteolytic enzymes from pollen and dust mites, as well as enterotoxin from Clostridium perfringens, attack junctional membrane proteins, which results in paracellular leakage (Runswick et al., 2007; Sonoda et al., 1999; Wan et al., 1999). In addition, several TJ transmembrane proteins function as receptors for viruses. For example, claudin 1 functions as a co-receptor for the hepatitis C virus and is required for virus entry (Evans et al., 2007). Similarly, several of the TJ-associated members of the CTX-protein family, such as the coxsackievirus and adenovirus receptor (CAR) and JAM-A (which binds to reovirus), also act as viral receptors (Barton et al., 2001; Cohen et al., 2001; Walters et al., 2002). In some cases (e.g. hepatitis C virus), binding of virus to the TJ protein favours its entry into cells, whereas in other cases the interaction helps to overcome the junctional diffusion barrier to enable the virus to access its actual receptor (e.g. rotavirus) or to promote the release of virus from the epithelium (e.g. adenovirus) (Evans et al., 2007; Nava et al., 2004; Walters et al., 2002). Another striking example is provided by the bacterium Helicobacter pylori, which causes gastric ulcers and cancer (Pritchard and Crabtree, 2006). H. pylori translocates a protein called CagA into host cells. CagA associates with the ZO-1–JAM-A complex, which is thought to contribute to corruption of the gastric epithelial barrier (Amieva et al., 2003). Because the binding of CagA causes the redistribution of ZO-1 complexes, it is possible that ZO-1-associated signalling mechanisms contribute to the development of H. pylori-induced pathologies.

A complex network of TJ-associated proteins has been identified over the last 20 years. Despite this, we have only begun to understand how these proteins interact with each other and cooperate to perform various junctional functions. Thus, the identification of specific TJ-associated molecular mechanisms remains a major challenge for the future. For example, many junctional proteins can bind to several other proteins and it seems unlikely that they associate with all of them at once; it will therefore be important to identify the role of particular interactions for junction assembly, and to determine when they occur and for which functions of TJs they are important. Moreover, substantial evidence indicates the existence of mechanisms that permit the paracellular permeation of specific ions and small hydrophilic molecules; however, the underlying molecular structure and machinery are only poorly understood and there is little structural information. Only a few animal models that are deficient in specific TJ proteins have thus far been generated and most have not been analysed in great detail. Moreover, TJ junctions will have to be studied in different species in the future, because the example of ZO-3 has shown that deficiency of a specific protein can lead to different phenotypes in different animal models (Adachi et al., 2006; Kiener et al., 2008; Xu et al., 2008). To understand the role of TJ-associated signalling mechanisms and junctional barrier properties, it will be crucial to determine the role of TJ-associated proteins in development and their importance for normal organ function. Finally, TJ proteins appear to contribute to many diseases, ranging from viral infections to cancer. The understanding of TJs at a molecular level, and of how they contribute to such diseases, should provide therapeutic opportunities for a wide range of disorders.

We apologise to anyone whose work and publications could not be cited due to the limited available space. M.S.B. and K.M. are supported by the Medical Research Council, the Biotechnology and Biological Sciences Research Council, the Association for International Cancer Research, Fight for Sight and the Wellcome Trust. The electron micrograph of the junctional complex in kidney epithelial cells is reprinted from Huber et al. (Huber et al., 1998), with permission. The electron microscopic image of a freeze-fracture replica from intestinal epithelial cells was kindly provided by Peter Munro, UCL Institute of Ophthalmology, London, UK.