Tight junctions consist of many proteins, including transmembrane and associated cytoplasmic proteins, which act to provide a barrier regulating transport across epithelial and endothelial tissues. These junctions are dynamic structures that are able to maintain barrier function during tissue remodelling and rapidly alter it in response to extracellular signals. Individual components of tight junctions also show dynamic behaviour, including migration within the junction and exchange in and out of the junctions. In addition, it is becoming clear that some tight junction proteins undergo continuous endocytosis and recycling back to the plasma membrane. Regulation of endocytic trafficking of junctional proteins may provide a way of rapidly remodelling junctions and will be the focus of this chapter.
An introduction to tight junctions
Tight junctions were first identified in electron micrographs, which revealed close contacts linking adjacent epithelial cells . These junctions, which are also found in endothelial cells, provide a barrier that regulates diffusion through the paracellular space (Figure 1). The history, function and molecular components of the tight junctions has been reviewed recently [2,3]. In the present chapter we will provide a brief introduction to tight junctions before focusing on their dynamic nature, and in particular the endocytic recycling of individual tight junction proteins.
Although tight junctions provide a barrier to diffusion, this barrier is not absolute as it allows the passage of certain solutes . In addition the permeability of tight junctions varies between different epithelial cells so that some epithelia form a ‘tight’ barrier whereas others are ‘leaky’. Tight junctions not only provide a barrier, but there is growing evidence they act as a platform for cell signalling molecules that regulate cellular behaviours, including gene expression and cell proliferation rates .
Molecular studies have shown that tight junctions consist of a large number of proteins, including members of the claudin, occludin and JAM (junctional adhesion molecule) families of transmembrane proteins . The claudin proteins, consisting of over 20 members in humans, are thought to be the main mediators of the epithelial permeability barrier [5⇓–7]. There are many lines of evidence to support this hypothesis, for example claudins can promote tight junction strand formation when expressed in fibroblasts  and mouse knockout studies show alterations in barrier function in a wide range of tissues, including the skin for claudin-1 knockouts  and kidney for claudin-2 knockouts . Importantly, some claudins promote a high-resistance barrier to diffusion, whereas others are thought to form pores that allow ionic diffusion . Thus variations in the expression profile of claudins is proposed to be a major determinant of tissue-specific variations in permeability . The other transmembrane tight junction proteins appear to have a role in the regulation/modulation of this claudin-based barrier [2,3]. A range of cytoplasmic proteins can associate with the transmembrane proteins, such as ZO (zonula occludens)-1 and cingulin, which indirectly link tight junctions to the actin cytoskeleton and various signalling proteins [11⇓–13]. The formation of these large multi-subunit structures allows the tight junctions to fulfil the various functions associated with them, including barrier formation and regulation of cell signalling.
The dynamic nature of tight junctions
Tight junctions are found in tissues that show dynamic behaviours. For example, in the mammalian intestine, epithelial cells are replaced every 4–5 days , and during pregnancy there is a massive increase in mammary epithelial cells which are then removed at the end of lactation . During the addition and removal of cells, tight junctions must be able to maintain barrier function. Tight junctions must also be remodelled as immune cells transmigrate through epithelia and during wound healing . Many stimuli, including inflammatory cytokines and bacterial infection, also rapidly modify the tight junction barrier . These observations suggest that rapid alteration of tight junction composition occurs and there is general acceptance that tight junctions, like the epithelial tissues they are found in, are dynamic structures. The dynamic behaviour of the tight junctions includes movement of tight junction strands and of individual tight junction proteins, which can diffuse within the tight junctions and in and out of the junctions [2,3]. There also appears to be continuous endocytosis and recycling back to the plasma membrane of tight junction proteins in a range of epithelial cell types [17–19], an aspect of tight junction dynamics that will be the focus of the present chapter.
Is continuous recycling of tight junction proteins a common feature of epithelial cells?
Following endocytosis from the plasma membrane into EEs (early endosomes), transmembrane cargo proteins can follow the degradative pathway to the lysosome or be recycled back to the plasma membrane. They can also undergo retrograde transport to the TGN (trans-Golgi network) where they may undergo further sorting . In epithelial cells the picture is further complicated by endocytosis from distinct apical and basolateral membrane domains into apical or basolateral EEs and transcytosis between the two domains [4,21] (Figure 2). There are also multiple independent pathways responsible for recycling proteins back to the plasma membrane  making endocytic protein trafficking, particularly in epithelial cells, an extremely complex process.
Identifying the trafficking fate of an endocytosed transmembrane tight junction protein can be achieved using a biotinylation assay that labels extracellular lysine residues of proteins allowing their trafficking to be monitored . The biotinylation assay was used to provide the first evidence that tight junction proteins can be continuously recycled in confluent epithelial cells when occludin was found to be endocytosed and recycled back to the plasma membrane in a mouse mammary cell line (MTD1A) . In contrast, claudin-1 was not endocytosed in these cells. However, a potential limitation of this study is that the rate of degradation of occludin was not measured. Owing to the nature of the biotinylation assay used it is possible that at least some of the occludin trafficking classified as recycling could have been a consequence of degradation.
Subsequent experiments with MDCK (Madin–Darby canine kidney) II cells, a canine kidney line  which is commonly used to study tight junctions, examined the trafficking of claudin-1 and occludin. Importantly, this work added an additional control so that degradation and recycling could be distinguished. Claudin-1 was found to be constantly endocytosed and recycled in MDCK II cells, with no detectable degradation observed over the time frame of these assays (20 min for degradation and recycling). Over longer time periods (hours/days) claudin-1 is degraded , indicating that a small percentage of endocytosed claudin-1 is directed for degradation in MDCK II cells. This would produce a gradual turnover of claudin-1 protein. Subsequent work showed that, like claudin-1, claudin-2 is endocytosed and recycled in MDCK II cells . In contrast, significant endocytosis of occludin  and claudin-4  was not detected in MDCK II cells using the same assay and incubation time (1 h), showing that tight junction proteins have different rates of flux through the endocytic system.
Analysis of additional epithelial cell lines showed that claudin-1 is constantly recycled in the human colon-cancer-derived line CaCo-2 and the human-lung-derived line 16-HBE . Interestingly, unlike in MDCK II cells, in these lines occludin was endocytosed. Its fate was then split, with some protein being recycled and the rest being degraded. Combining results from the four different epithelial cell lines currently analysed shows that recycling of at least a subset of tight junction proteins is a common feature of epithelial cells. In addition there is cell-type-specific variation in the rate at which individual tight junction proteins are trafficked through the endocytic system. This data is summarized in Table 1. Most of this work is very recent and there are a number of issues that remain to be addressed. For example, it will be important to determine which other tight junction proteins are recycled and whether similar rates of recycling occur in vivo. It is also important to stress that the biotin assay does not label cytoplasmic proteins. This means it cannot be used to follow the fate of proteins such as ZO-1, which could be transported with the transmembrane tight junction proteins they associate with.
How is endocytosis of tight junction proteins mediated?
The first step in endocytic trafficking is the internalization of transmembrane proteins from the plasma membrane. Endocytosis can occur via clathrin-dependent  and clathrin-independent mechanisms, which include caveolae-driven endocytosis and macropinocytosis . Endocytosis of tight junction proteins has been reported to occur by all of these mechanisms: for example calcium depletion, a non-physiological stimulus that induces tight junction breakdown, triggers endocytosis of occludin, claudin-1 and JAM-A by a clathrin-mediated pathway . Studies on stimulus-induced remodelling have revealed clathrin-independent internalization of tight junction proteins. For example, occludin endocytosis occurs via caveolae following Escherichia coli CNF-1 (cytotoxic necrotizing factor-1) stimulation  or TNFα (tumour necrosis factor α) stimulation . Recent work also shows that the chloride channel CIC-2 acts to reduce caveolae-mediated endocytosis of occludin in CaCo-2 cells . Claudin-1, occludin and JAM-A are internalized via macropinocytosis in IFNγ (interferon γ)-treated T84 cells . Live imaging studies using fluorescent fusion proteins have revealed a further feature of endocytosis for claudin-3 . This involves a peculiar mechanism whereby plasma membranes from two juxtaposed cells are internalized into one of the cells. This has been referred to as ‘eat-each-other’ endocytosis.
It should also be noted that different tight junction proteins are not always endocytosed en masse, but can be endocytosed independently of one another. For example, EGF (epidermal growth factor) stimulation of MDCK II cells increases endocytosis and degradation of claudin-2 without affecting claudin-1 . In addition, occludin endocytosis occurs in latrunculin-A-treated MDCK II cells  and in vivo in anti-CD3-treated mice  without concomitant internalization of claudin proteins. The ability to independently endocytose proteins may allow cells to alter the repertoire of tight junction proteins at the plasma membrane and fine-tune paracellular permeability in response to changing conditions.
An interesting question is what determines the selection of tight junction proteins as endocytic cargo, and one possibility is that post-translational modification of individual tight junction proteins is responsible. Ubiquitination, phosphorylation and palmitoylation are reversible modifications that have all been linked to endocytosis . Transmembrane tight junction proteins have also been shown to undergo these modifications [35,36]. Expression of the ubiquitin ligase LNX1p80 drives endocytosis of claudin-1 , whereas an alternative ubiquitin ligase Itch promotes endocytosis of occludin . An attractive hypothesis is that low level activity of these, and possibly other ubiquitin ligases, promotes continuous endocytosis of tight junction proteins. Regulation of specific ubiquitin ligase activity could also alter endocytic rates of individual tight junction components. However, knockdown or knockout experiments ablating the function of these proteins and analysis of the effect on endocytosis rates are required to establish if this is the case. Phosphorylation of claudins, occludin and JAMs have all been reported, with the phosphorylation status having an affect on tight junction localization [38⇓⇓–41]. For example, phosphorylation of Ser490 of occludin is associated with its endocytosis and barrier loss in endothelial cells . Finally, claudin family proteins contain conserved signature cysteine residues that are palmitoylated and are required for efficient localization at tight junctions . Palmitoylation is a reversible protein modification that has been linked to endocytosis  in addition to other protein trafficking events [21,45,46] making it a potential regulator of claudin endocytosis.
Another important issue is where endocytosis of tight junction proteins occurs. It is plausible that tight junction proteins are not endocytosed from the junctions themselves, but from adjacent (or even distant) regions of the plasma membrane. Live imaging and computer modelling experiments have been performed to investigate the dynamics of individual tight junction components in living cells . These studies indicate that occludin within the tight junctions continuously exchanges with an extra-tight junction pool. Furthermore, claudin-1 is located in the lateral membrane in addition to the junctional complexes of polarized epithelial cells [48⇓–50]. Cytoskeletal rearrangements at tight junctions are likely to be crucial in facilitating junctional remodelling and endocytosis. If endocytosis occurs away from the tight junctions, then regulation of the rate that tight junction proteins move out of junctions and into regions of the plasma membrane undergoing active endocytosis would provide another possible mechanism of regulating endocytic cargo selection. The mechanisms, pathways and modifications mentioned are not mutually exclusive, and a complex interplay is likely to regulate selection for endocytosis.
Regulation of recycling and junctional remodelling
While many studies have considered stimulated endocytosis as an important process in tight junction remodelling, relatively little attention has focussed on post-endocytic sorting of tight junction components. However, the continuous recycling of junctional proteins [17–19] raises the possibility that a reduction in the rate of recycling could achieve the same goal as increased endocytosis, that is, removing tight junction proteins from the plasma membrane and causing either an accumulation of internal protein or increased degradation. The post-translational modifications, ubiquitination, phosphorylation and palmitoylation, mentioned above, all have roles to play in post-endocytic sorting , in addition to endocytosis, and so provide potential signals to regulate the rate of recycling of tight junction components. This leads to the question of whether control of post-endocytic sorting, in particular recycling, participates in junctional remodelling.
Evidence to suggest that control of post-endocytic sorting is important comes from studies on the fate of tight junction proteins in IFNγ-treated epithelial cells [30,51]. Occludin, claudin-1 and JAM-A all accumulate intracellularly, in compartments containing RE (recycling endosome) markers Rab4 and Rab11, following IFNγ treatment, with no increase in degradation of these proteins. Removal of IFNγ results in release of the accumulated proteins back to tight junctions, so it is reasonable to postulate that IFNγ treatment blocks recycling. Similarly, other manipulations of epithelial cells, such as removal of calcium  or incubation with CNF-1 , reversibly displace tight junction proteins to intracellular locations, indicating that regulation of recycling may be a common mechanism by which tight junction composition can be controlled. Intriguingly, following CNF-1 treatment, occludin does not co-localize intracellularly with claudin or JAM-A, providing evidence that these proteins could be recycled by distinct routes. Control of tight junction recycling is not limited to epithelial cells as claudin-5 and occludin recycle in endothelial cells following recovery from chemokine treatment .
Post-endocytic sorting and trafficking of tight junction proteins
In order to understand how stimuli might modify post-endocytic trafficking it is important to establish the endocytic routes taken by tight junction proteins and identify the proteins responsible for this trafficking. Following endocytosis the EEs are the primary destination for material removed from the plasma membrane. Proteins entering these endosomes must be sorted to degradative, recycling or retrograde endocytic trafficking pathways and this sorting is a major determinant of the fate of endocytosed proteins .
A group of proteins which are tightly linked to post-endocytic sorting is the ESCRT (endosomal sorting complex required for transport). This complex has a well-established role in the trafficking of ubiquitinated transmembrane proteins to lysosomes for degradation [53,54] and is important for attenuating signalling from growth factor receptors such as the EGFR (EGF receptor) . However, in addition to blocking the degradative pathway, inhibiting ESCRT function can cause defects in the recycling of receptors, such as those for EGF, transferrin, asialoglycoprotein and low-density lipoprotein [56⇓⇓–59]. Experiments using a dominant-negative ESCRT protein showed that ESCRT function is required for the continuous recycling of claudin-1, and when ESCRT function is inhibited claudin-1 accumulates intracellularly . Why is ESCRT function required for claudin-1 recycling? ESCRT proteins are known to interact with deubiquitinating enzymes, and the timing of deubiquitination is emerging as an important factor in modulating the fate of endocytosed proteins [60,61]. Therefore it is possible that perturbations in the ESCRT machinery leads to mis-regulation of ESCRT-associated deubiquitinating enzymes resulting in incorrect post-endocytic sorting and a failure in claudin-1 recycling.
Addition of YM201636, a small molecule inhibitor of the lipid kinase PIKfyve (FYVE finger-containing phosphoinositide kinase) also inhibits claudin-1 and claudin-2 recycling . This kinase is responsible for the synthesis of PtdIns(3,5)P2 from PtdIns3P on early endosomes and its function has been linked to a number of endocytic sorting events [62,63]. It is unclear why PIKfyve function might be required for recycling of claudins, but as PtdIns(3,5)P2 binds the ESCRT III component Vps24 (vacuolar protein sorting 24)/CHMP3 in vitro , it is possible that recycling is blocked by inhibiting the recruitment of the ESCRT machinery to endosomes.
Members of the Rab family of small GTPases have well-documented roles in vesicular trafficking , and experiments using a dominant-negative construct showed that Rab13 is required for the continuous recycling of occludin in MTD1A cells . siRNA (small interfering RNA) knockdown experiments have also shown that the trafficking of internalized claudin-1 to the plasma membrane after a calcium switch (calcium depletion and subsequent repletion) requires Rab13 and its binding protein JRAB/MICAL-L2 . This identifies Rab13 as a key mediator of tight junction recycling, although it also has an additional role in the biosynthetic delivery of cargoes from the TGN to endosomes in polarized epithelial cells . Rab11 is well known to be involved in recycling, and tight junction proteins accumulate in Rab11-positive compartments following stimulation with IFNγ or CNF-1 [28,30], but we are not aware of any studies that have investigated the requirement of Rab11 in tight junction protein recycling. The ESCRT, PIKfyve and Rab13 studies described above have identified the first few proteins whose function is required for continuous tight junction recycling. However, it is clear that much more research is needed to provide a full description of the molecular basis for these events and to elucidate how specificity of recycling of individual tight junction proteins is achieved.
Why do cells continuously recycle tight junction proteins?
The final, and perhaps most important, question is why should cells expend the energy required to constantly move tight junction proteins into the cell and then back to the cell surface? A simple explanation is that no system is perfect and as some proteins may be internalized inappropriately, recycling would provide an efficient system to return these to the plasma membrane. Tight junction recycling may also have evolved as a consequence of the dynamic nature of tight junctions. If a cell needs to form more tight junctions it can reduce endocytosis and/or increase the rate of recycling to increase the amount of tight junctions at the cell surface. Conversely, if the area of tight junctions needs decreasing a cell can promote endocytosis and/or reduce recycling to remove excess tight junction proteins from the plasma membrane. This provides a flexible way to rapidly deal with physiological variations in tight junction size. It could also be used to fine-tune barrier function: for example, the composition of claudins at the plasma membrane could be altered to allow cells/tissues to regulate paracellular permeability.
The recycling of tight junctions may also be important in pathological conditions. Alterations of tight junctions have been associated with tumour formation  and blocking recycling would reduce the amount of functional tight junction protein at the plasma membrane, perhaps reducing the stability of epithelial tissues. Modulation of tight junctions also occurs in inflammatory bowel disease and following bacterial infection [16,28,30,51]. The evidence described above suggests that inflammatory cytokines and bacterial proteins may act, in part, by blocking tight junction recycling. These potential links to disease illustrate the importance of future work aimed at fully understanding this process.
The demonstration that tight junction recycling appears to be a common feature of epithelial cells throws up many questions such as how is it controlled and why is it important? Our understanding is still at a rudimentary level: tight junction recycling is known to occur in several different types of epithelial cells and the first few proteins which are required for this process have been identified, but many more experiments are required to provide a true mechanistic understanding. Key issues that need to be addressed include (i) what proteins mediate trafficking through the endocytic system, (ii) is there more than one tight junction recycling pathway and (iii) whether this process is regulated. Rates of recycling in vivo also need to be determined and, finally, the question of whether tight junction recycling is altered during pathological conditions such as cancer, inflammatory bowel diseases and bacterial infection needs to be addressed.
• Tight junctions provide a permeability barrier that is dynamic and constantly undergoing remodelling.
• A number of tight junction components, including claudins and occludin, are constitutively endocytosed and recycled in epithelial cells.
• There is cell-type-specific variation in the tight junction proteins that are recycled.
• ESCRT, PIKfyve and Rab13 are required for tight junction recycling.
• Tight junction recycling may be aberrantly regulated in disease.
Work in our laboratory on this area was supported by a Cancer Research UK project grant [grant number C26932/A9548]. We thank Dr Christopher Caunt (University of Bath) for critical comments on the draft of the chapter.
- © The Authors Journal compilation © 2012 Biochemical Society