Ruffles and Spikes: Control of tight junction morphology and permeability by claudins

Abstract

Epithelial barrier function is regulated by a family of transmembrane proteins known as claudins. Functional tight junctions are formed when claudins interact with other transmembrane proteins, cytosolic scaffold proteins and the actin cytoskeleton. The predominant scaffold protein, zonula occludens-1 (ZO-1), directly binds to most claudin C-terminal domains, crosslinking them to the actin cytoskeleton. When imaged by immunofluorescence microscopy, tight junctions most frequently are linear structures that form between tricellular junctions. However, tight junctions also adapt non-linear architectures exhibiting either a ruffled or spiked morphology, which both are responses to changes in claudin engagement of actin filaments. Other terms for ruffled tight junctions include wavy, tortuous, undulating, serpentine or zig-zag junctions. Ruffling is under the control of hypoxia induced factor (HIF) and integrin-mediated signaling, as well as direct mechanical stimulation. Tight junction ruffling is specifically enhanced by claudin-2, antagonized by claudin-1 and requires claudin binding to ZO-1. Tight junction spikes are sites of active vesicle budding and fusion that appear as perpendicular projections oriented towards the nucleus. Spikes share molecular features with focal adherens junctions and tubulobulbar complexes found in Sertoli cells. Lung epithelial cells under stress form spikes due to an increase in claudin-5 expression that directly disrupts claudin-18 / ZO-1 interactions. Together this suggests that claudins are not simply passive cargoes controlled by scaffold proteins. We propose a model where claudins specifically influence tight junction scaffold proteins to control interactions with the cytoskeleton as a mechanism that regulates tight junction assembly and function.

1. Introduction

A major epithelial function is to provide a barrier that separates two distinct microenvironments, the apical and basolateral compartments of a wide range of organs. To support a physiologically functional barrier, epithelial cells must be selectively permeable to ions and solutes. Selective permeability requires cells to regulate two different pathways across the epithelial barrier: the transcellular and the paracellular routes that occur through and between cells, respectively.

Paracellular transport is regulated by specialized intercellular points of contact that form the apical junctional complex (AJC), which separates polarized cells into distinct apical and basolateral domains. The AJC encircles each cell, pairing with neighboring cells to create an adhesive network formed by several classes of intercellular junctions, including adherens junctions, tight junctions, gap junctions and desmosomes [1, 2]. The AJC also establishes the apical/basolateral polarity axis by organizing the Crumbs and Partitioning defective complexes [3]. The multifunctional nature of the AJC enables intercellular communication (gap junctions), provides mechanical integrity to epithelial monolayers (adherens junctions and desmosomes) and acts as a signaling hub that is sensitive to cell contact through differential interactions between transmembrane and cytosolic junction proteins [4]. In addition, the AJC also serves as a site for recruitment and organization of the actin cytoskeleton [1, 5].

Tight junctions are the AJC component that regulates paracellular barrier permeability to water, small molecules, and ions (Figure 1). The main determinants of tight junction-regulated paracellular permeability are claudin-family transmembrane proteins. Claudins form paracellular ion channels of varying specificity and permeability (reviewed in [6–8]). Tissue-specific claudin composition allows for organ-specific paracellular permeability. Claudin composition and assembly into tight junctions is also sensitive to environmental stressors, such as inflammation. Moreover, claudins do not act in isolation. In concert with other transmembrane proteins, including other claudins, MarvelD proteins (e.g. occludin, tricellulin) and Ig-superfamily proteins (e.g. JAM-A), claudins form complexes with cytoplasmic scaffold proteins that regulate interactions with the actin cytoskeleton. In addition to their role as paracellular channels, there is increasing evidence that claudins can also serve as part of a signaling hub through their specific interactions with different classes of scaffold proteins [9, 10].

Figure 1. Protein composition of tight junctions and adherens junctions.

Shown is a subset of transmembrane, cytosolic scaffold and cytoskeletal proteins associated with tight junctions (occludin, claudin, ZO-1, ZO-2) and adherens junctions (cadherin, α-catenin, β-catenin).

In addition to the regulation of ion and water permeability, tight junctions also regulate the paracellular flux of soluble molecules, including large macromolecules [11]. Soluble molecules do not move through stable, claudin-based pores. Instead, their diffusion across tight junctions is due to transient discontinuities that create a path of diffusion [12, 13]. Tricellular junctions also form a path for paracellular diffusion of soluble molecules that is regulated independently from bicellular tight junctions [14, 15]. Here, we consider changes to the morphology of bicellular tight junctions that correlate with increases in paracellular permeability.

One implication of the ability of claudins to differentially recruit tight junction scaffold proteins is that changes in claudin composition can impact scaffold/cytoskeletal interactions, thereby affecting the overall organization of tight junctions. This can be recognized by two characteristic non-linear tight junction morphologies that we refer to here as “tight junction ruffles” and “tight junction spikes”. Tight junction ruffles (Figure 2b) are largely parallel to the site of cell-cell contact but they differ from linear tight junctions (Figure 2a) in that they deviate from the most direct path interconnecting tricellular contact sites. By contrast, tight junction spikes are structures that are perpendicular to tight junctions along sites of cell-cell contact (Figure 2c). As indicated in Figure 2 and described in detail below, linear tight junctions, ruffles and spikes are associated with characteristic differences in the organization of junction associated actin filaments. In addition to tight junction ruffles and spikes, non-continuous distributions of claudins (e.g. strand breaks and puncta) at cell-cell contact sites also can influence paracellular permeability. Ruffles, spikes and strand breaks all correlate with impaired paracellular barrier function and thus provide valuable indicators of altered assembly of tight junction proteins.

Figure 2. Roles for actin in control of tight junction morphology.

A. Linear tight junctions showing cortical actin and symmetrical forces perpendicular to the plane of the membrane (k₁ = k₂). B. Tight junction ruffles, with tight junctions tethered to actin perpendicular to cortical actin and subjected to higher, symmetrical forces than linear junctions. C. Tight junction spikes subjected to asymmetrical tension (k₁ > k₂). and oriented along actin stress fibers.

In this review, we describe signal transduction events that induce changes in claudin composition driving changes in tight junction morphology to regulate barrier function. We propose a model where interactions between claudins, scaffold proteins, and the actin cytoskeleton alter tight junction morphology and function by influencing the balance of tension at intercellular junctions.

2. Ruffled junctions

When imaged by immunofluorescence microscopy, tight junctions typically appear as a relatively straight, continuous line that connects tricellular contact points (Figure 3), however, there are several conditions where tight junctions exhibit a ruffled morphology [11, 16, 17]. Ruffled tight junctions have been observed for several years (e.g. [18, 19]). More recently they were systematically quantified by Tokuda et al. [20] in a study correlating changes in claudin expression by MDCK cells with differences in the extent of tight junction ruffling.

Figure 3. Quantitation of tight junction ruffles.

A. ZO1 in HIF1β deficient Caco2 cells has a ruffled appearance. Transfection to overexpress claudin-1 cDNA normalizes ZO1 distribution to a linear morphology. B. Quantification of tight junction ruffling was performed by dividing the actual junction length (dotted line A) by the distance between tricellular junctions (dashed line B). Examples of ruffled (left) and linear (right) tight junction morphology are shown. Reproduced from [27] with permission.

Other terms used to describe ruffled tight junctions include: wavy [21–23], tortuous [20, 24–26], undulating [18, 27], serpentine [11, 26] or zig-zag [20, 28]. Referring to these structures as tight junction ruffles parallels the term plasma membrane ruffles, formed by the leading edge of migrating cells [29]. In addition to comparable morphology, the mechanisms that drive plasma membrane ruffles at the leading edge and tight junction ruffles are likely to be comparable, (e.g. actin reorganization and branching by factors such as WASP) [30].

To date there have not been any examples of other junction proteins showing a ruffled morphology. Although there are no a priori reasons why other classes of junctions (e.g. adherens junctions) could not assume a ruffled conformation, junctional ruffles are likely unique to tight junctions. For instance, E-cadherin localization is not ruffled in intestinal epithelial cells that have tight junction ruffles [31].

Ruffled junctions have a distinct appearance (Figure 3) and can be quantified by a measure sometimes referred to as the “zig zag index” [20]. The zig zag index is the actual path length of a tight junction between two tricellular junctions (A) divided by the minimum path length (B). A junction is considered ruffled if A/B is significantly larger than 1, where 1 is a completely unruffled (or linear) tight junction.

Tight junction ruffling frequently correlates with increased paracellular permeability (or leak) [27, 32], although that is not always the case [20]. One intriguing hypothesis is that ruffling increases permeability by increasing tight junction circumference, thus enabling more functional claudin channels per cell [24]. In addition, ruffled and linear tight junctions are differentially associated with actin which is also likely to have an impact on their barrier function [23].

Many stimuli have been shown to induce ruffling, including molecular manipulation of tight junction proteins, impaired oxygen signaling, integrin-mediated signaling and direct mechanical stimulation. Examples of each of these stimuli and the impact they have on claudin composition and tight junction morphology are described below and in Table 1.

Table 1 –

Stimuli inducing ruffled tight junctions

Stimulus	Effect on	Effect on	Effect on	Reference
	Claudins	TER	paracellular flux
High expression of ZO-1	cldn-2 high, cldn-1, cldn-7 low	Significant change	Variable degrees of changes in permeability, but no real pattern	Tokuda, et al. [20]
ZO-1 truncation mutants	nd	nd	nd	Fanning et al. [19, 43]
TOCA-1 expression	No change in cldn-2	No significant change	Increase 3kDa Dextran	Van Itallie, et al. [54]
KD HIF1B knockdown	Decrease cldn-1	Decrease	Increase FITC dextran (3, 10, 40 kDa)	Saeedi, et al. [27]
KD HIF-2a knockdown	nd	Decrease	nd	Glover, et al. [31]
Reoxygenation after anoxia injury	Increase in cldn-4	Decrease	Increase FITC-dextran	Jin, et al. [21]
MLCK activation	Local decreases in cldn-1	Decrease	Increase inulin, mannitol	Shen, et al. [18]
Cyclic stretch	nd	nd	Increase FITC inulin	Samak, et al. [23]
VAV3 inactivation	nd	Decrease	nd	Hilfenhaus, et al. [60]
Nanostructure contact	Decrease cldn-1, cldn-4	Decrease	Increase FITC-BSA, FITC-IgG, Etanercept	Kam, et al. [16], Walsh, et al. [17], Stewart, et al. [11]

2.1. Roles for claudin/ZO-1 interactions in tight junction ruffling

Claudins interact with each other both across tight junctions (trans-interactions) and within tight junctions (cis-interactions) [33–35]. In addition the claudin C-terminal cytoplasmic domain interacts with cytosolic scaffold proteins, which crosslink these proteins to the cytoskeleton and can also act as a signaling hub [34, 36, 37]. Foremost among these is the tight junction scaffold protein zonula occludens-1 (ZO-1), which has a PDZ1 domain that binds to the “YV” motif found at the extreme C-terminus of most, but not all claudins [38]. Other proteins that interact with the claudin YV motif include ZO-2 and ZO-3 [39], as well as other non-ZO related proteins such as the E3 ubiquitin ligase LINXp80 and COPII cargo sorting protein Sec24C, both of which have been shown to play a role in regulating incorporation of claudin-1 into tight junctions via vesicular trafficking [40, 41].

ZO-1 helps crosslink claudins to the actin cytoskeleton [19] and is uniquely implicated in the control of junction ruffling. This was demonstrated in MDCK II cells where ZO-1 depletion or low levels of ZO-1 resulted in tight junctions that were highly linear, whereas high levels of ZO-1 expression were associated with significant tight junction ruffling [20].

MDCK II cells engineered to be deficient in five claudins (MDCK quinKO) show non-ruffled, linear ZO-1 labeling under the control of JAM-A, underscoring a need for claudins in the formation of ruffled junctions [42]. Tight junction ruffling is unique to ZO-1/claudin interactions, since knocking out or overexpressing ZO-2 or ZO-3 has little effect on tight junction morphology [20]. Moreover, in order for ZO-1 to induce tight junction ruffles, it needs to have both the actin binding motif as well as the U6 region of the GUK domain [43]. Interestingly, the ZO-1 U6 domain plays a key role in conformational shifts in ZO-1 that limit occludin binding [43, 44]. This further supports a model where ZO-1 binding to claudins, but not occludin, form more ruffled junctions in contrast to the linear tight junctions produced with ZO-1 binding concurrently to claudins and occludin.

When MDCK II cells are transduced to overexpress ZO-1, the increase in tight junction ruffling is also associated with an increase in tight junction-associated claudin-2 [20]. Consistent with a role for claudin-2 in regulating tight junction ruffling, MDCK I cells, which express low levels of claudin-2, tend to have less ruffled tight junctions than MDCK II cells that express high levels of claudin-2 [20, 45]. Claudin-2 is a pore forming claudin that increases tight junction ion and water permeability [46, 47]. Ruffled junctions have a higher capacity for claudin-2, which likely further enhances this effect [24].

Claudin-2 competes with other claudins for the ability to integrate into tight junctions, including claudin-1, claudin-4 and claudin-7 [20, 48, 49]. Although claudin-2 is less efficiently assembled into tight junction strands than claudin-1 and claudin-4 [50], claudin-2 has a longer half-life [51] and thus remains more effectively associated with tight junctions as compared with claudins having a shorter half-life. Control of claudin-2 turnover is a function of the C-terminal domain and does not require ZO-1 binding, suggesting that other, as yet unknown, factors uniquely regulate claudin-2 integration into tight junctions [51].

Although high levels of claudin-2 correlated with tight junction ruffling, MDCK II cells deficient in claudin-2 expression did not have fully linear tight junctions [49]. Instead, increased expression of other claudins is also required to fully linearize tight junctions. For instance, claudin-2 deficient MDCK II cells transduced with exogenous claudin-4 have more linear tight junctions than claudin-2 deficient cells alone [50]. The ability of other claudins to influence formation of ruffled or linear tight junctions will require screening them for their effect on tight junction morphology and permeability.

How claudin-2 influences tight junction ruffling remains to be determined, although evidence is emerging that different claudins can influence downstream interactions between ZO-1 and other scaffold proteins. For instance, ZO-1 enhances assembly of claudin-1 into tight junction strands through interactions with the PDZ1 and PDZ3 motifs of ZO-1, whereas, claudin-2 assembly requires the PDZ1 and PDZ2 motifs [52]. Potential roles for the ZO-1 PDZ2 motif in claudin-2 recruitment into tight junctions include the PDZ2 motif mediating ZO-1 dimerization [53] or binding to other scaffold proteins. As one possibility, claudin-2 may promote folding of ZO-1 into a conformation that promotes binding of the F-BAR protein TOCA-1 complexed to WASP, leading to termination of branched actin filaments at junctions [54] (Figure 4).

Figure 4. Model for claudin-directed changes in ZO-1 conformation.

A. Claudin-1 binds to ZO-1 in a conformation enabling interactions with occludin that promote association with actin in a cortical orientation, parallel to the plane of the plasma membrane. B. ZO-1 associated with claudin-2 is proposed to have an alternative conformation. Shown here are induced interactions with TOCA-1 (crescent) and WASP (red bar), potentially re-orienting actin/ZO-1 interactions into a conformation that favors tight junction ruffling.

Claudin-dependent switching of ZO-1/scaffold protein complexes also provides a potential mechanism where the orientation of actin filaments interacting with tight junctions can switch between cortical (parallel to the plane of the plasma membrane) and filamentous (roughly perpendicular to the plasma membrane) (Figure 2). In this model, the tension exerted on ruffled tight junctions is higher than linear junctions, yet still symmetrical across the plane of the junction.

It is well established that myosin light chain kinase (MLCK) and rho family kinases regulate barrier function by altering the magnitude of tension on tight junctions [1, 55, 56]. Differential tension can also lead to changes in ZO-1 conformation that affect its function and ability to interact with other proteins, including claudins [57]. In addition to tension, flow can also impact barrier function. For instance, blood flow through veins is much slower than through arteries, and veins are considerably more permeable than arteries [58, 59]. Consistent with this difference in permeability, venous endothelial cells have more ruffled junctions and are associated with actin stress fibers as opposed to arterial endothelial cells that form high resistance barriers and have linear junctions associated with cortical actin [60].

Taken together, this suggests a model where claudin-directed reorientation of the actin cytoskeleton coordinated with changes in actomyosin-mediated tension regulates tight junction morphology and barrier function. Consistent with this model, tight junction ruffling was reversed by treatment with the myosin inhibitor blebbistatin, further underscoring a role for actin-associated tension in ruffle formation [20].

2.2. Hypoxia induced tight junction ruffles

Epithelial barrier function is highly sensitive to changes in oxygen tension, where each epithelial tissue has a particular oxygen set point ranging from hyperoxia (high oxygen tension) to hypoxia The lung is an example of a hyperoxic tissue whereas the intestine and, counterintuitively, skin are hypoxic [61–63].

Oxygen tension is sensed by the Hypoxia Inducible Factor (HIF)-1α and HIF-2α (Endothelial PAS Domain Protein 1; EPAS1) transcription factors that act in concert with HIF-1β [64, 65]. At normoxia, prolines on HIF transcription factors become hydroxylated targeting them to the proteasome to be degraded. However, in hypoxia, the non-hydroxylated forms of HIF-1α and HIF-2α translocate to the nucleus where they activate gene transcription.

Although HIF-1α and HIF-2α activate different subsets of the genome (e.g. [31]) both influence epithelial tight junctions, since depletion of either of these proteins experimentally or due to chronic inflammation impairs barrier function [66, 67]. Specifically, it has been demonstrated in human intestinal epithelial cell lines that knockdown of either HIF-1α [27] or HIF-2α [31] induces a ruffled tight junction morphology as determined by immunofluorescence as well as decreased barrier function.

Despite the comparable effects of shRNA knockdown on tight junction morphology and permeability, HIF-1α and HIF-2α have different mechanisms of action. HIF-1α is directly linked to claudin-1 expression, since HIF-1α knockdown in intestinal and esophageal epithelial cells decreases claudin-1 and reporter assays demonstrate that HIF-1α interacts with the CLDN1 promoter [27, 68]. HIF-1β depleted cells show reduced claudin-1 expression (because of the impact on HIF-1α) and increased tight junction ruffling. Critically, transducing HIF-1β depleted cells to overexpress claudin-1 reverses the ruffled tight junctions into a linear morphology and restores barrier function, indicating a direct role of claudin-1 in regulating paracellular permeability that corresponds with tight junction assembly [27].

In contrast to HIF-1α, HIF-2α does not directly regulate claudin-1 transcription [68], despite the observation that HIF-2α knockdown also induces tight junction ruffling. Instead, HIF-2α depletion decreases expression of several key enzymes involved in creatine metabolism, including creatine kinase M (CKM) and creatine kinase B (CKB), enzymes that otherwise co-localize with E-cadherin and ZO-1 [31]. Critically, creatine supplementation rescues intestinal epithelial barrier function of HIF-2α deficient cells in vitro and in a dextran sodium sulfate inflammatory bowel disease model in vivo, underscoring a role for localized energy metabolism in regulating tight junction morphology and function. It remains to be determined whether CK and claudin-1 overlap or represent parallel pathways that regulate the extent of tight junction ruffling.

While increasing claudin-1 expression leading to increased barrier function is due in part to the barrier forming properties of claudin-1 [69], the precise mechanisms whereby claudin-1 changes tight junction morphology have not been fully elucidated. As described above, the influence of claudin-1 on ZO-1 function can affect the recruitment of other proteins that can then affect tight junction morphology. However, with the exception of ZO-1, specific claudin-1 interacting proteins that determine whether tight junctions are ruffled or linear have not yet been identified.

2.3. Integrin stimulation by nanostructured surfaces

Contact of the basal surface of cells with the extracellular matrix has a considerable impact on cell phenotype and function, which is a key element in the ability to produce organoid cultures that faithfully mimics differentiated cell behavior in native tissues [70].

Specifically, receptors known as integrins bind to extracellular matrix components regulating the organization of the actin cytoskeleton that, in turn, have several downstream consequences impacting cell function [71]. In addition to the native biological substrates for integrins, recent work has determined that integrin contact with synthetic, nanostructured surfaces alters epithelial barrier function in a geometry-dependent manner [16, 32]. The effects of nanostructured surfaces on cells depend on several parameters, including feature aspect ratio, density, pattern and substrate chemistry [32].

Several classes of nanostructured surfaces imprinted on inert polymers have been shown to increase paracellular permeability through direct contact with β1 integrin [16, 17]. This has utility for design of devices for transdermal delivery of macromolecular therapeutics (e.g. Etanercept), since coating microneedles with a nanostructured surface significantly enhances macromolecule delivery as compared with bare stainless steel microneedles by increasing keratinocyte transepithelial permeability [17]. Agents delivered transdermally via nanostructure coated microneedles also are more effectively delivered to the cardiovascular and lymphatic systems by an as yet unknown mechanism [17, 72, 73]. One possibility that remains to be tested is that dermal cells stimulated by nanostructure contact secrete factors promoting downstream vessel permeability.

Epithelial cell contact with specific nanostructured surfaces increases paracellular leak and causes junctions to become ruffled [11, 16, 17]. This is accompanied by decreased expression of claudin-1 [17], consistent with the effect of HIF-1α knockdown described above. Claudin-4 expression is also reduced by nanostructure contact, which may be directly associated with an effect of nanostructures on integrins, since claudin-4 is closely associated with β1 and α2 integrin [74].

In addition to the effects on claudin expression, nanostructure contact also stimulates focal adhesion kinase (FAK) and MLCK activity, both of which were required for the increase in paracellular permeability [16, 17]. Whether the changes in claudin expression and kinase activation have an additive or redundant effect on tight junction morphology is not yet known.

2.4. Ruffles formed by mechanical stimulation

Mechanical stimulation of cells can also lead to tight junction ruffling and changes in paracellular permeability. A particularly dramatic example of this is cyclic stretch of Caco-2 cells [23]. Cyclic stretch activates MLCK, suggesting a potential mechanism comparable to the effect of nanostructured surfaces on cells. Cyclic stretch also activates JNK and Src, which phosphorylate ZO-1 and occludin [23] and are likely to influence their ability to interact with each other (e.g. [75–77]) and potentially other proteins. Consistent with the effects of mechanical stress on tight junction assembly, precision cut lung slices subjected to stretch caused dissociation of claudins from ZO-1 in lung epithelial cells [78]. Moreover, cells transduced with constitutively activated MLCK show regions of localized ruffling that are deficient in claudin-1, further underscoring a role for claudin-1 in maintaining linear tight junctions [18].

3. Tight junction spikes and discontinuities

In contrast to tight junction ruffles, tight junction spikes are an asymmetric deviation from linear tight junction morphology. Tight junction spikes appear as projections at cell-cell interfaces that orient in a perpendicular direction from junctions towards the nucleus (Figure 5). The asymmetry of tight junction spikes is shared by a comparable adherens junction structure, focal adherens junctions, that also can be asymmetric and have been studied in considerable detail (reviewed in [79, 80]). Several other terms have been used to describe focal adherens junctions [81], including: perpendicular junctions [82, 83], spot junctions [84], discontinuous junctions [83, 85], punctate junctions [83], junction-associated intermittent lamellipodia [86] and buttons [87]. A comparable structure formed by desmosomes has been referred to as linear arrays [88] and another formed by gap junctions has been referred to as filadendrites [89].

Figure 5. Tight junction spikes induced in lung epithelial cells.

Alveolar epithelial cells isolated from alcohol or control-fed rats were cultured for 7 days on Transwell permeable supports and immunolabeled for claudin-18. Cells from alcohol fed rats showed enhancement of tight junction spikes, that are claudin-18 projections perpendicular to the cell-cell interface (arrowhead). Square regions in the top panels correspond to magnified images below. Note strand breaks, puncta and other discontinuities in claudin-18 present in cells from alcohol-fed rats (Bar, 10 μm). Reproduced from [108] under CC BY 4.0.

Here we distinguish tight junction spikes from clearly discontinuous tight junctions [83, 87, 90–92], in that spikes typically project from intact regions of intercellular tight junctions.

While visually distinct, tight junction discontinuities and spikes also are quantifiable by image analysis of the relative amount of continuous, punctate and perpendicular junctions [83, 93], using segmentation image analysis [94] or by neural network analysis of patterns of junctional disruption based on differential labeling intensity [92].

Tight junction spikes differ from focal adherens junctions which are usually punctate. Also, tight junction spikes formed by alveolar epithelial cells are clearly distinct from adherens junctions, since they are deficient in the cadherin-binding protein β-catenin, which instead is localized to areas that are adjacent to areas where tight spikes are formed [95]. The punctate nature of focal adherens junctions may reflect dissolution of lateral cadherin interactions that are weaker than trans cadherin interactions and thus more easily disrupted by increased tension [96, 97]. Another key difference is that tight junction spikes more likely form from mature tight junctions as opposed to focal adherens junctions that tend to be precursors to fully mature adherens junctions [79, 82].

Tight junction discontinuities generally correlate with gross disruption of the actin cytoskeleton [55, 98] leading to paracellular leak. By contrast, tight junction spikes align with actin filaments perpendicular to intercellular tight junctions [95, 99, 100]. Actin also has a comparable role in organizing spikes formed by desmosomes [88] and gap junctions [89].

Although tight junction ruffles and spikes are both organized by actin filaments that are perpendicular to the plane of the plasma membrane, they differ in that ruffles are organized by comparable, symmetric actin filaments on both sides of the AJC, however the arrangement of actin in spikes is asymmetric (Figure 2). Also, spikes are organized along the actin filaments (much as linear junctions are aligned along cortical actin) whereas ruffles are tethered to them. Otherwise, the molecular mechanisms that underlie tension generation and induces ruffle and spikes are comparable (e.g. MLCK, Rho kinase activation) [55, 56]. Several other molecular features are conserved between ruffles and spikes, including recruitment of vinculin [17, 82] and F-BAR proteins [54, 81] as regulators of cytoskeletal tension and membrane curvature, respectively.

3.1. Tight junction spikes as organizers of vesicular traffic

It has long been appreciated that formation of adherens junctions precedes tight junction formation [101]. This has previously been associated with the relative strength of trans interactions between cadherins as opposed to claudins. A more subtle role for adherens junctions in stabilizing tight junctions was revealed by an examination of α-catenin-deficient EpH4 epithelial cells, which were subject to constitutive delivery and endocytosis of claudin-3 to the plasma membrane [102]. The inability of α-catenin-deficient cells to form tight junctions was not due to a loss of mechanical junction stability, but instead was linked to an imbalance in plasma membrane cholesterol content. Replenishing cell cholesterol re-established the assembly of claudin-3 into tight junctions and stimulated the formation of claudin-3 containing spikes that also contained cholesterol [102]. These findings are consistent with previous studies demonstrating that tight junction proteins preferentially partition into cholesterol enriched microdomains [103] but extend this observation to include spikes as well as established tight junctions.

Although tight junctions appear to be relatively stable structures, in fact they are highly dynamic and are readily endocytosed [104–107]. In cells subjected to oxidative stress, tight junction spikes serve as active “hot spots” for vesicle budding and fusion [108]. Moreover, Eph4 epithelial cells plated at low density form tight junction spikes at cell-cell interfaces between two cells migrating in opposite directions; these spikes show double membrane structures by electron microscopy, indicating that one cell endocytoses both halves of a tight junction [104]. These data suggest that tight junction spikes are associated with responses to cell stress and/or tension. Whether spikes reflect unique vs. constitutive processes that regulate tight junction turnover is an open question at present.

Tight junction spikes are reminiscent of a structure found in seminiferous tubule junctions, the basal tubulobulbar complex [109]. Tubulobulbar complexes are enriched in claudin-11, which has a limited pattern of expression and may be uniquely required for their formation [110]. Tubulobulbar complexes are enriched for actin, actin-binding proteins, dynamin and are active sites of vesicle budding and fusion, all of which are associated with tight junction spikes in other epithelial cells.

Interestingly, tubulobulbar complexes are also associated with endoplasmic reticulum-plasma membrane (ER-PM) contact sites, which form a calcium signaling-complex that controls junction remodeling [111]. A comparable ER-PM contact site is also involved in epidermal growth factor receptor (EGFR) endocytosis and signaling [112]. It also has been shown that in MDCK II cells, EGFR specifically induces claudin-2 endocytosis, but not claudin-1 endocytosis [113]. Whether claudin-2 turnover induced by EGFR occurs by a spike-mediated pathway is not known at present.

Claudin endocytosis is a regulated process. Moreover, different claudins are internalized by different endocytic pathways [105], which provide mechanisms to regulate barrier function by differential regulation of endocytosis. For instance, claudin-1, claudin-2 and claudin-4 are internalized by clathrin-mediated endocytosis, however claudin-5 is preferentially internalized by caveolar endocytosis [105, 107, 113]. Since claudins form complexes, it is likely that lateral claudin-claudin interactions can influence the endocytic pathways that mediate claudin turnover [33, 35, 114].

Stimulation of acinar epithelial cell mAChR with carbachol induces claudin-4 phosphorylation, resulting in formation of a complex with β-arrestin2, subsequent internalization of claudin-4 and loss of barrier function [107]. Inhibiting clathrin-mediated endocytosis prevented the loss of claudin-4 and preserved barrier function. Involvement of tight junction spikes in this process was revealed by treatment with the proteasome inhibitor MG132, which stabilized spike-associated claudin-4 and also preserved barrier function.

3.2. Spikes formed in response to chronic alcohol exposure are due to impaired claudin/ZO-1 interactions

Chronic alcohol abuse is a risk factor for poor outcome in acute respiratory distress syndrome [115, 116]. This is due, in part, to the deleterious effect of alcohol exposure on lung epithelial barrier function [117]. Increased paracellular leak across alveolar epithelial cell monolayers is accompanied by an increase in tight junction spikes [108] (Figure 5). The effects of alcohol on alveolar epithelial tight junctions, including increased leak and stimulation of spike formation, can be recapitulated by TGFβ1 [99] and antagonizing GM-CSF [95], indicating that alcohol causes an imbalance in lung epithelial cytokine signaling.

Claudin-18 is prominently expressed by alveolar epithelial cells however, the healthy lung epithelium expresses low levels of claudin-5 [118]. In response to alcohol exposure, alveolar epithelial cells increase claudin-5 expression, which correlates with an increase in tight junction spikes containing claudin-18 [108]. Increased claudin-5 expression was both necessary and sufficient to induce spikes in alveolar epithelial cells. Using super-resolution microscopy and the proximity ligation assay to measure protein-protein interactions in situ, it was determined that increased claudin-5 binds to claudin-18 and inhibits it from interacting with ZO-1, resulting in increased tight junction spike formation [108] (Figure 6).

Figure 6. Model for claudin-claudin interactions affecting scaffold protein binding.

A. Tight junctions enriched for claudin-18 show significant binding with ZO-1, as well as other associated proteins, indicated by the blue square, and actin in a cortical orientation (equivalent to Figure 4A). 18. B. Increased claudin-5 interacts with claudin-18 to prevent an interaction with ZO-1. The red oval and grey circle denote putative C-terminal interacting proteins that bind to claudin-18 in the absence of ZO-1. In this model, claudin-5 is proposed to induce a conformational change in the C terminal domain of claudin-18 (arrows).

Although the precise mechanism by which claudin-5 affects claudin-18/ZO-1 interactions remains to be determined, it seems likely that there will be other examples of claudin-claudin interactions that affect organization of the tight junction scaffold. One possible model is that claudin-5 binding to claudin-18 causes a conformational shift in the C-terminus of claudin-18 displacing ZO-1 and enabling other, as yet unknown, factors to interact with claudin-18 (Figure 6). Whether this is the case will require identifying proteins that preferentially interact with spike associated claudin-18.

3.3. Roles for claudins in regulating tight junction ultrastructure

Tight junctions have been examined at the ultrastructural level, using freeze fracture scanning electron microscopy, demonstrating a diversity of tight junction organization as meshworks that differ in strand number, shape and organization. By and large, tight junction permeability inversely correlates with meshwork depth and strand number (e.g. [119–121]) although this is not always the case [122]. Tight junction ruffles do not necessarily correlate with changes in ultrastructure since there are examples where ruffled junctions do [43] and do not [18] have accompanying changes in tight junction ultrastructure that can be detected by freeze fracture electron microscopy.

Claudins are required to form tight junction strands at the ultrastructural level [42, 123] and the architecture of the tight junction meshwork is sensitive to claudin composition. For instance, overexpression of claudin-3 by MDCK cells causes a transition from an angular to a curved loop meshwork structure and decreased strand breaks [124]. The third transmembrane domain of claudin-3 has a unique bent conformation that has been directly linked to the control of tight junction strand morphology by altering claudin packing [125]. Increased claudin-4 expression by MDCK cells produces tight junctions that have a reticular network of parallel strands, whereas high levels of claudin-2 expression are associated with curved stands that are diffuse [122].

Imaging using conventional confocal immunofluorescence microscopy has a limit of resolution of 200 nm. This is not sufficient resolution to detect strand breaks in the range of 20 nm - 200 nm, which are associated with increased paracellular leak due to changes in claudin expression [124, 126]. Super-resolution fluorescence microscopy has the capacity to image tight junction strands at high enough resolution to reveal differences in the ultrastructural meshwork formed by different claudins; this was demonstrated by analysis of claudin-null HEK293 cells transfected to express claudin-3 or claudin-5, which showed differences in tight junction ultrastructure that could be detected by freeze fracture electron microscopy and Spectral Position Determination Microscopy [127]. In native alveolar epithelial cells, tight junction spikes were detected by stochastic optical reconstruction microscopy (STORM) [108]. However, alveolar epithelial cells are squamous and have a limited tight junction meshwork architecture [128, 129], so STORM did not detect any meshwork changes associated with tight junction spikes. Using super-resolution microscopy to assess ultrastructural changes formed by native claudins in cuboidal epithelia is feasible using current technology, but likely challenging, since it will require super-resolution in the x-z axis in addition to the x-y plane.

4. Summary and future directions

Tight junction assembly and function are influenced by protein composition, post-translational modifications and the internal and external mechanical forces they are subjected to. Most models emphasize the impact of actin and the cytosolic scaffold on the assembly and behavior of claudins. However, evidence is emerging that this is a reciprocal relationship, where claudins themselves can be active determinants of scaffold protein conformation and function.

Claudins associated with ruffles are assembled into tight junctions. However, it is not known whether claudins associated with tight junction spikes are assembled into bona fide tight junctions. Cells forming tight junction spikes show evidence that intact tight junctions are maintained when they were engulfed by one cell from another [104, 108]. However, it is also possible that spikes contain a pool of non-junction associated claudins. One method to distinguish whether spike associated claudins are fully integrated into tight junctions is to use Fluorescence Recovery After Photobleaching (FRAP) analysis of YFP-tagged claudins which can differentiate junction associated claudins from non-junctional pools, based on rate and extent of recovery [48]. If spike associated claudins are not junctional, they could serve other roles. For instance, non-junctional pools of claudin-7 along the lateral plasma membrane regulate tumor cell growth and migration [130, 131].

Since most approaches to measure epithelial permeability are based on overall measurements of an intact monolayer or tissue, the impact of tight junction morphological changes on paracellular permeability have not been well elucidated. Electrophysiologic methods that rely on scanning live cell monolayers to map local paracellular ion permeability have been developed, although these are difficult to use and correlate with tight junction morphology because they are low throughput [132, 133].

Several imaging approaches have been established that enable local permeability to be measured. This includes a fluorescence barrier permeability assay based on plating cells on a biotinylated substrate that are subsequently probed with fluorescently tagged streptavidin and imaged by fluorescence microscopy (XPerT assay) [134]. The XPerT assay has been successfully used to identify sites of localized barrier dysfunction, primarily in endothelial cell monolayers [78, 135–137]. The ZnUMBA assay based on zinc permeability and a fluorescent reporter molecule represents another approach to visualize localized barrier permeability [138]. Coupling imaging methods with cells expressing fluorescently tagged tight junction proteins will enable sites of paracellular leak to be identified relative to areas where tight junctions are not linear.

Many advances have been made in defining the tight junction proteome, including the use of BioID to identify proteins that are in close proximity to ZO-1, claudin-4 and occludin [74, 139]. The utility of this approach is underscored by the finding that the N- and C- terminal domains of ZO-1 interact with different proteins [139]. Further expanding the use of BioID to identify proteins that interact with other claudins comparing conditions where tight junctions are linear, ruffled or forming spikes are anticipated to help define mechanisms where claudins control tight junction morphology and could help identify new proteins specific to ruffled or spike morphologies.

The ability of claudins to influence their own assembly and integration into tight junctions is beginning to be appreciated. Claudin-1, claudin-2 and claudin-5 have been associated with linear, ruffled and spiked tight junctions respectively. The ability of other claudins to influence tight junction morphology is less well established. In addition, the effect of claudins on tight junction morphology is likely to be context sensitive, especially due to interactions with other claudins present in tight junctions, and remains to be determined.

Undoubtedly, C-terminal domains of different claudins bind to different protein substrates, however, evidence is now emerging that claudins can influence the behavior of scaffold and other proteins. By analogy with connexins [140–142], the C-terminal domains of claudins are likely to be intrinsically disordered having significant structural plasticity. ZO-1 also has intrinsically disordered domains, is mechanosensitive and can exist in different phase states [143], underscoring the concept that tight junction assembly is highly context dependent with respect to both local protein composition and biophysical mechanical state. Taken together, we propose a model where complexes between different claudin C-terminal domains and scaffold proteins influence each other to fold into unique conformations. One implication of this model is that determining the regulation of epithelial paracellular barrier function will require taking into account how the reciprocal interplay between claudins, scaffold proteins and cytoskeletal tension affect tight junction assembly and function.

Highlights.

• Tight junctions show hallmark morphological changes, ruffles and spikes, in response to stimuli that cause paracellular leak.

• Ruffling and spike formation are due to claudin-directed regulation of tight junction scaffold protein engagement of the actin cytoskeleton.

• Tight junction ruffling requires ZO-1

• Symmetric and asymmetric changes in tension preferentially induce ruffles and spikes, respectively.

Abbreviations

AJC

apical junctional complex

BAR

Bin/Amphiphysin/Rvs

EGFR

epidermal growth factor receptor

FRAP

Fluorescence Recovery After Photobleaching

GUK

guanylate kinase

HEK

Human Embryonic Kidney

HIF

hypoxia induced factor

mAChR

muscarinic acetyl choline receptor

MDCK

Madin Darby Canine Kidney

MLCK

myosin light chain kinase

PDZ

postsynaptic density protein (PSD95), Drosophila disc large tumor suppressor (DlgA), and zonula occludens-1 protein (ZO-1)

STORM

STochastic Optical Reconstruction Microscop

TOCA

Transducer of Cdc42 dependent actin assembly

WASP

Wiskott–Aldrich syndrome protein

XPerT

express micromolecule permeability testing

ZnUMBA

Zinc-based Ultrasensitive Microscopic Barrier Assay

ZO

zonula occludens