Non-canonical functions of claudin proteins: Beyond the regulation of cell-cell adhesions

ABSTRACT

Tight junctions form a barrier to the diffusion of apical and basolateral membrane proteins thus regulating membrane polarity. They also regulate the paracellular movement of ions and water across epithelial and endothelial cells so that functionally they constitute an important permselective barrier. Permselectivity at tight junctions is regulated by claudins, which confer anion or cation permeability, and tightness or leakiness, by forming several highly regulated pores within the apical tight junction complex. One interesting feature of claudins is that they are, more often than not, localized to the basolateral membrane, in intracellular cytoplasmic vesicles, or in the nucleus rather than to the apical tight junction complex. These intracellular pools of claudin molecules likely serve important functions in the epithelium. This review will address the widespread prevalence of claudins that are not associated with the apical tight junction complex and discuss the important and emerging non-traditional functions of these molecules in health and disease.

KEYWORDS:

• claudin
• EMT
• nuclear transcription
• review

Introduction

Tight junctions have a conserved 3-D structure of which our current knowledge has advanced considerably as described in numerous recent reviews.^1-4 Tight junctions were initially described from electron micrographs as the fusion of lateral membranes from the most apical part of adjacent cells with no intercellular space present over a variable distance.⁵ By freeze-fracture replica electron microscopy, this interesting structure was further shown to consist of a band of anastomosing intramembrane strands that run horizontal to the cell surface and confer either leakiness or tightness to the epithelium.^6,7 In conventional electron micrographs, tight junctions were shown to contain numerous “kisses” at the plasma membrane of adjoining cells that interacted with the actin cytoskeleton.⁸ These structures were later found to correlate with highly regulated transmembrane pores made by claudin molecules.³ Through this system of paracellular pores that are regulated by a set of interacting transmembrane and accessory proteins linked to the actin cytoskeleton, tight junctions were thus determined to regulate paracellular ion transport across the epithelial sheet.^1,9 The nature of this complicated membrane structure also conferred a “fence” function that allowed for the sorting of apical and basolateral membrane proteins above and below the tight junction, respectively. The claudin family of tight junction proteins currently consists of 27 known members,¹⁰ with good evidence to suggest that they form the tight junction strands visualized by freeze-fracture replica electron microscopy.⁴ Considerable evidence exists to support the notion that cells express numerous claudin molecules simultaneously, which are chosen on a function-based paradigm to regulate paracellular ion and water permeability.

The general architecture of apical tight junctions is the same in epithelial cells from many organs including intestine, kidney, and lung.^11-15 Sertoli cells of the testes also have a classical tight junction structure, which in part helps to create the blood-testis barrier.¹⁶ Although the epidermis consists of a stratified, rather than a simple epithelium, classical tight junctions form in one of the distinct cellular layers to regulate paracellular permeability and ion transport.¹⁷ Endothelial cells, particularly those forming the blood-brain barrier, also have classical tight junction architecture.¹⁸ Even some tissues without cells that have an apical and basolateral polarity form tight junctions to regulate ion permeability, including cardiac muscle cells in the heart and myelinated neurons.^19-22

Despite a considerable focus on the role of apical tight junctions in regulating paracellular ion transport and membrane polarity, mounting evidence suggests that tight junction claudin proteins are localized to sites outside of the tight junction complex. Tight junction transmembrane and accessory proteins, such as occludin and junction adhesion molecule (JAM-A) are not usually present in these “extra-tight junction” sites, suggesting that claudins serve important functions outside of their classical role in regulating ion permselectivity. Thus, the aim of this article is to review our present knowledge about the location and function of these “extra-tight junction” claudin proteins and to highlight their emerging role in regulating cellular signaling, cell adhesion, epithelial to mesenchymal transformation (EMT), cell migration/invasion/metastasis, and the transcription of genes that, at least in part, regulate cell survival and proliferation. Rather than categorizing the extra-tight junction claudins as “mislocalized” tight junction proteins, we suggest naming them differently. This might include tCLDN for tight-junction associated claudins, bCLDN for claudins localized to the basolateral membrane, and cnCLDN for claudins that shuttle between the cytoplasm and nucleus. Creating a new designation for these proteins may be necessary to fully consider their important functional role outside of the tight junction complex in health and in disease.

Basolateral membrane-associated claudins and cellular signaling

When freeze fracture replica electron microscopy of stomach epithelial cells was done in 1973 by Goodenough and Claude,⁷ discontinuous tight junction-like structures (strands) were found along the basolateral membrane that were somewhat dismissed because they were thought to play no role in generating the transepithelial electrical resistance. Because claudin molecules constitute at least one component of tight junction strands,⁴ these results suggested that claudin proteins are localized at the basolateral membrane in addition to the tight junction. Immunolocalization studies confirmed the basolateral localization of some claudin molecules in stomach (Fig. 1) and similar results were obtained in many different tissues (Table 1).

Figure 1. Confocal micrographs of the stomach mucosa stained for tight junction associated claudin-4 and basolateral membrane-associated claudin-18. Note the absence of green fluorescence signal at apical tight junctions in the surface of tissues stained for claudin-4 but the strong signal at apical tight junctions of epithelial cells in the neck and base (arrows). In contrast, claudin-18 is highly concentrated at the basolateral membrane of all epithelial cells (arrowheads) in the stomach mucosa with particularly robust expression in surface epithelial cells. MM, muscularis mucosa.

Table 1. Tissues and cultured cells in which tight junction proteins are found outside the apical tight junction complex.^*

Tissue/Cultured cell	Claudin	Species	Localization	Reference(s)
Skin
Epidermis	1,4,7	Human	Basolateral membrane	^17
	2,5	Human	Cytoplasmic	^17
	1,4	Mouse	Basolateral membrane	^84
	23	Mouse	Cytoplasmic	^17
Epididymis
	1	Rat	Basolateral membrane	^27
Testis
	3	Mouse	Strong basal plus lateral membrane below the tight junction	^85
Stomach
Antrum	1	Human	Basolateral membrane	^86
Corpus	1	Human	Basolateral Membrane	^87
All segments	3	Rat	Basolateral membrane	^23
All segments	5	Rat	Basolateral membrane	^23
Corpus	18	Canine, Human, Mouse	Basolateral Membrane	^35,88-90
Intestine
Small intestine (mouse, Jejunum; rat, all segments rat)	2	Mouse, Rat	Basolateral membrane, crypts only	^23,25
Small intestine (all segments)	3	Rat	Increasing crypt to surface basolateral membrane staining (weak basolateral expression overall)	^23
Small intestine (all segments)	4	Mouse, Rat	Basolateral membrane, villus only	^23
Small intestine (all segments)	5	Rat	Basolateral membrane (weak)	^23
Small intestine (all segments)	7	Mouse	Basolateral membrane	^24
Small intestine (ileum)	8	mouse	Basolateral membrane (exclusively)	^24
Large intestine (all segments)	2	Rat	Basolateral membrane (weak), crypts only	^23
Cecum	3	Mouse	Basolateral membrane (strong)	^25
Large intestine (all segments-rat)	3	Rat	Increasing crypt to surface basolateral membrane staining (strong basolateral expression overall)	^23
Large intestine (all segments)	4	Rat	Basolateral membrane, villus only	^23
Large intestine (all segments)	5	Rat	Basolateral membrane, crypt only (weak)	^23
Colon	1	Human	basolateral membranes and cytoplasmic granules	^26
Colon	2	Human	Surface: basolateral membrane; Crypts: basolateral membrane and cytoplasm.	^26
Colon	4	Human	Basolateral membrane, surface and crypt	^26
Colon	7	Human	Basolateral (mostly basal), highly expressed in surface cells and weakly expressed in crypt cells	^91
Colon	7	Mouse	Basolateral membrane	^24
Colon	8	Mouse	Basolateral membrane (exclusively)	^24
Colon	13	mouse	Basolateral membrane with strong expression on basal membrane	^24
Kidney
Nephron (distal convoluted tubule)	7	Mouse	Basolateral membrane	^92
Nephron (distal convoluted tubule, collecting tubule, thick ascending limb, loop of Henle)	7	Rat, Pig	Basolateral membrane	^93
Loops of Henle and collecting ducts	7	Mouse	Basolateral membrane	^92,94
Lung
Airway epithelium	1, 4	Human	Basolateral membrane	^95
Basal cells	1,4	Human	Surrounds cells	^95
Airway epithelium	7	Human	Basolateral membrane (exclusively)	^95
Cultured Cells
Caco-2 cells (Intestinal)	1, 7 EpCAM	Human	Basolateral membrane	^39
	15	Human	Nucleus	^76
Keratinocytes	9, 23		Cytoplasmic	^17
MDCK (strain II, Kidney)	1 Occludin	Canine	Basolateral membrane	^96
	ZO-1, ZO-2	Canine	Nucleus	^61,62
T84 cells (Colonic)	1, 7, EpCAM	Human	Basolateral membrane	^39

Note.

* The localization may also be found associated with the apical tight junction complex.

One of the first demonstrations that claudin molecules localize to sites other than the apical tight junction complex in tissues was by Rahner et al,²³ who showed that claudins can be expressed along the basolateral membrane of gastrointestinal (GI) epithelial cells and additionally that multiple claudins co-localize at these sites (Table 1). Studies in mouse and human GI tissues confirmed and extended this initial finding (Table 1).^24-26 In rat epididymis, the expression of basolateral claudin-1 was even more interesting because some tissue segments showed solely lateral localization below the tight junction, some segments showed very strong basal versus lateral localization, and one segment showed extremely strong apical localization specific to progenitor cells (basal cells) that sit along the basement membrane and never reach the epididymal surface (Table 1).²⁷ These results suggest that the lateral and basal membrane compartments have specific functional requirements that include claudin molecules. The important roles currently identified for basal claudins will be discussed in the next section.

One pertinent question to ask is “why is there a pool of basolateral membrane claudins in the first place”? Although little is currently known about the role of basolateral membrane claudins, it is possible that they represent a pool of claudin molecules that are available to recycle to the apical tight junction complex when needed (Fig. 2). It is well-established that the apical tight junction complex is highly dynamic and undergoes continual remodeling,²⁸ using a microtubule-based transport system to carry-out continuous endocytic recycling,^29,30 which is largely clathrin-mediated.³¹ Apically-directed tight junction proteins are sorted in the trans-Golgi network (TGN), and targeted to the apical tight junction complex by specific trafficking proteins (Fig. 1A).^32,33 In contrast, virtually nothing is known about how basolateral claudin molecules are sorted or targeted to the basolateral membrane and whether they follow conventional or canonical pathways.³⁴ Despite the vast literature on regulation of apically-targeted tight junction protein formation, trafficking, recycling, and turn-over, little evidence exists to support that a basolateral pool of claudins is needed for remodeling the apical tight junction (Fig. 2). Perhaps we do not have the appropriate tools currently available to evaluate membrane recycling from this pool.

Figure 2. Schematic diagram of potential functions for basolateral claudin molecules. (A) Basolateral claudins may function in the trafficking of endosomal vesicles (red circles) from the baolateral membrane to tight junctions for rapid remodeling. (B) Basolateral claudins may function as accessory pores to facilitate further discretion of ion transport at tight junctions (dotted arrow) and/or create signaling hubs via interaction with claudin/claudin complexes and the action cytoskeleton. EpCAM, epithelial cell adhesion molecule; TGN, trans-Golgi network.

Figure 3. Schematic diagram of the cell/extracellular matrix interface in epithelial cells. Claudin molecules interact with integrin -α and -β subunits within the focal adhesion complex, which provides and anchor for cells to the extracellular matrix. The presence of claudins results in MAPK signaling through the focal adhesion kinase (FAK)/Src/p 130cas complex.

Three other possibilities, at the least, could explain the function of basolateral membrane-associated claudins. First, basolateral membrane claudins may function as an accessory structure to the tight junction, serving as an additional permeability barrier to increase the transit time of charged molecules (Fig. 2). This notion is supported by studies using CLDN18A2.1 (stomach isoform)-deficient mice, which express no stomach-specific claudin-18 at the basolateral membrane and show a faster rate of H⁺-diffusion,³⁵ suggesting that claudin 18A2.1 normally forms channels or pores along the basolateral membrane to impede paracellular cation movement. Second, it is possible that claudins are able to diffuse freely along the basolateral membrane and are trapped into tight junction fibrils as they oligomerize within the apical tight junction complex. Although this type of membrane trapping has been described for exchange between basolateral and junctional pools of E-cadherin,^36-38 it has not yet been studied for claudins within the apical tight junction complex. Lastly, basolateral claudins may function as a signaling hub, signaling complex, or cluster, which integrates, codifies, and transports information into the cell (Fig. 2). This possibility has the strongest supporting data. To form a signaling complex or cluster along the basolateral membrane, claudin molecules should interact with one another and be immobilized in specific domains rather than diffusing randomly in the plane of the membrane (Fig. 2). Such an interaction occurs in T84 and Caco-2 intestinal cells, which express both claudin-1 and -7 along the basolateral membrane (Table 1).³⁹ It was recently shown that these 2 claudins additionally interact with and bind to epithelial cell adhesion molecule (EpCAM), forming a functional unit along the basolateral cell membrane.³⁹ This interaction was required to stabilize claudins at the basolateral membrane and was also required to regulate their rate of endocytosis and subsequent lysosomal degradation.³⁹ The interaction with EpCAM occurred with basolateral claudins and not with apical tight junction-associated claudins,³⁹ which can be retained within the tight junction complex by phosphorylation and not trafficked to and degraded by lysosomes.⁴⁰ Overall, these data suggest that cells expressing EpCAM, or other yet undefined claudin-binding partners, organize claudin molecules into discrete functional units along the basolateral membrane. Furthermore, these structures may be the source of discontinuous strands along the basolateral membrane that were originally identified by freeze fracture replica electron microscopy.⁷ Although this possibility is likely, the interpretations are not yet fully supported by data. With little known in this area of tight junction biology, a significant opportunity exists to determine novel claudin binding partners and to evaluate their functional role along the basolateral membrane.

Basal membrane claudins-cellular/extracellular matrix interactions

The basal membrane of epithelial cells is adjacent to an extracellular matrix (ECM) that, with the assistance of cellular ECM receptors (integrins), regulates cell adhesion, migration, differentiation, and the survival of epithelial cells (Fig. 3).⁴¹ Integrins perform these functions by organizing a signaling complex consisting of focal adhesion kinase (FAK), Src, and p130cas that is held together by their association with accessory proteins like talin, paxillin, vinculin, α-actinin, and the actin cytoskeleton (Fig. 3).⁴¹

At least 3 claudin molecules, claudins-1, -2, and -7, have been implicated in the regulation of cell/ECM interactions by integrating with integrin molecules in focal adhesions (Fig. 3) and regulating the phosphorylation of FAK to ensure adhesion of epithelial cells to the ECM. This function is required for normal cellular homeostasis but can also be used by cancer cells to facilitate the adhesion of metastatic cells.

Recent studies demonstrated that claudin-7 and the integrin-β1 subunit (lung) or the integrin-α2 subunit (with claudin-1, intestine) form a macromolecular complex that restricts claudin-7 to focal adhesions along the basal membrane (Fig. 3).^42-44 Little is known about the configuration of claudins within the focal adhesion complex. The downregulation of claudin-7, by genetic silencing methods, disrupts adhesion to the extracellular matrix, dysregulates the phosphorylation of FAK, and results in the downregulation of integrins, cell cycle, cell survival, and cell proliferation markers.^42,44 To highlight the importance of this claudin-integrin interaction along the basal membrane, intestinal epithelial cells are unable to attach to the underlying mucosa in CLDN7-deficient mice in vivo, resulting in ulcerative lesions that promote inflammation and death shortly after birth.⁴² Overall, these results suggest that claudin-7 interactions are essential to stabilize cell/matrix interactions and to regulate cell proliferation and survival by integrating, in some way, with effectors in the focal adhesion complex. Although it was originally thought that claudin-7 per se may transcriptionally regulate the expression of numerous effectors, particularly the expression of cell proliferation and survival genes, recent genomic profiling and pathway analysis demonstrated that claudin-7 indirectly regulates gene transcription via a cell signaling network with MAPK's as the central node (Fig. 3).⁴³ Other claudins also show strong expression patterns along the basal membrane (Table 1), suggesting that they are also involved in interactions at the focal adhesion complex. Further work would be required to test the role of these additional basally-localized claudins in various cell types.

The localization of claudin-7 and its interaction with integrins in particular areas of a tissue (like crypt vs. villus in intestine) or segments of the same tissue (like proximal vs. distal vs. collecting ducts in the kidney) may be a key regulator of gene expression including the expression of tight junction-associated claudins, and thus of the permeability characteristics in different segments of the same tissue. One example to support this notion is in kidney, where proximal tubule epithelium, which is classified as a “leaky” epithelium due to robust claudin-2 expression, has low claudin-7 expression.⁴⁵ In the absence of β1 integrin, which was genetically deleted, the same cells changed phenotype to that of a “tight” epithelium as is found in the distal tubules and collecting ducts, which have minimal claudin-2 expression and high claudin-7 expression.⁴⁵ This conversion (claudin-2/claudin-7), driven by the absence of integrin β1, resulted in significant functional abnormalities in the kidney due to the lack of conventional proximal tubule function.⁴⁵ Although the relationship of claudin-7 expression and the expression of cell matrix effectors was not investigated,⁴⁵ it would be interesting to know whether or not high levels of claudin-7 expression stabilized the epithelium at focal contacts, activated MAPK or other signaling pathways, and transcriptionally regulated the expression of claudin-2 and other effectors that are constitutively expressed in distal tubule or collecting duct cells. These data however, suggest that cell signaling from focal adhesions on the basal membrane of epithelial cells, via an integrin-claudin pathway, could be a more generalized mechanism for conferring regional specificity to the mucosal barrier in a given tissue. More data, however, is needed to lend support to this interpretation.

Although integrin/claudin-7 interactions at the basal membrane are part of the story, claudin-7 is also known to interact directly with EpCAM along the basal membrane.⁴⁶ This association occurs in many tissues including colon and pancreas and in numerous cancers, for which the interaction is thought to mediate tumor progression.⁴⁷ Although there are several phosphorylation sites on claudin molecules that could regulate their activity, palmitolyation in particular regulates the ability of claudin-7 to interact with integrins, recruit EpCAM, and concomitantly associate with the actin cytoskeleton.⁴⁸ The histological features of intestinal epithelial cells in EpCAM-deficient mice look similar to CLDN7-deficient mice, most prominently with intestinal epithelial cells unable to attach to the underlying mucosa.⁴⁹ These results suggest that, at least for claudin-7, EpCAM interactions are essential for cell/ECM interactions that confer mucosal homeostasis.

In contrast to the important role claudin-7/integrin/EpCAM interactions play in normal mucosal function, the expression of claudin molecules in metastatic cells likely plays a detrimental role by facilitating cell/ECM interactions in cancer progression. One example is in breast cancer cells, where claudin-2 expression drives the cell surface expression of integrins α2β1 and α5β1, forming an integrin-claudin adhesion complex to facilitate the formation of liver metastases.⁵⁰ Perhaps the most novel use of claudins as adhesion receptors is in HIV infection, where the virus was shown to anchor claudin-7 to its envelop and use this protein to adhere-to and infect CD4(-) target cells.⁵¹ Thus, while the expression of claudins at the basal membrane function to regulate beneficial cell/ECM interactions and cell signaling, this important function can also be used to promote cellular adhesion in disease processes that are not beneficial such as the adhesion of metastatic cells to promote cancer progression or to promote HIV infection.

Basolateral membrane claudins-regulation of epithelial-mesenchymal transformation, cell migration, invasion, and tumorigenesis

EMT is defined as the transformation of polarized epithelial cells, which are attached to one another in the epithelial sheet by tight and adherens junction complexes, into individual cells with a migratory (spindle-shaped) phenotype that is accompanied by the loss of membrane-associated E-cadherin, the expression of mesenchymal markers including nuclear β-catenin, the activation of stemness genes including CD44, and the activation of numerous cell signaling pathways.⁵² EMT also occurs in wound healing and in tumorigenesis, allowing for the migration of epithelial cells. One interesting aspect of EMT is that although adherens junctions disappear, which should lead to the dissolution of tight junctions and the downregulation of tight junction protein expression, select claudin molecules are highly expressed. In nearly every article written about the role of claudins in EMT/wound healing and in tumorigenesis, the role of claudins in permeability at the apical tight junction complex is discussed when in fact the newly expressed claudin molecules are localized to the basolateral membrane, intracellular compartments, or the nucleus. Little is known about the role of claudins when they are expressed in these locations. While literature on the topic of claudins in EMT and cancer progression is rather extensive, the studies typically correlate changes in claudin expression to disease pathogenesis and/or patient outcome. The few mechanistic studies outlined below have shed light on the role of claudins in EMT and cell/ECM interactions, and cell migration/invasion in cancer pathogenesis.

Numerous studies have shown that claudins regulate the expression and activation of matrix metalloproteinases (MMT's),^53-56 which causally link claudins to ECM remodeling that is required for EMT, cell migration, and invasion in cancer pathogenesis. MMP's are secreted at the basolateral surface of epithelial cells and function to change the cell/ECM interaction to promote migration and invasion. Claudins have been shown to recruit membrane-type MMP's, enhancing the activation of pro-MMP, although the exact role of claudin molecules in this process is unknown.⁵⁷ In addition to matrix remodeling, MMP's have also been shown to activate important transcriptional regulators, like EGFR, which promotes cell signaling cascades that alter claudin expression.⁵⁸ Claudins may synergize to promote MMP activation, as was recently demonstrated by claudin-1/claudin-6 interactions in human AGS cells.⁵⁹ In that study, claudin-6 induced MMP-2 activation via claudin-1 expression, which accelerated the rate of cell migration and invasion.⁵⁹ Lastly, the invasiveness of cancer cells can be regulated by epigenetic means including DNA methylation, which modulates claudin expression by inhibiting (or stimulating) claudin gene promoter activity.⁶⁰

In in vivo studies, CLDN7-deficient mice showed robust MMP expression that accompanied the inability of cells to attach to the underlying mucosa,⁴² suggesting a direct correlation between matrix remodeling and cell/ECM attachment. It is not clear whether all claudins have this capability and as such, further work is needed.

Nuclear claudins as transcriptional regulators

The nuclear localization of tight junction proteins was first shown for ZO-1 and ZO-2 (Table 1),^61,62 and subsequently their role in gene transcription via interaction with ZONAB/DbpA or c-myc/E-box elements, respectively, was elucidated.^63,64 The presence of claudins in the nucleus was highlighted by Dhawan et al,⁶⁵ who showed that the expression of claudin-1 was localized to the nucleus in tissues from patients with human primary colon cancer, in colon cancer-derived liver metastases, and in metastasis to lymph nodes, whereas there was no nuclear claudin-1 expression in normal human colonic mucosa. Rather, claudin-1 was localized to the basolateral membrane in normal human colonic mucosa.⁶⁵ Colon cancer cell lines had differential claudin-1 expression, with SW480 and HCT116 cells expressing little to no claudin-1, respectively, and SW620 cells expressing high levels of claudin-1.⁶⁵ Immunostaining studies in the colonic cancer cell lines showed that the highly expressing line had strong nuclear localization of claudin-1, and cell fractionation studies consisting of nuclear, cytoplasmic, and membrane fractions validated that highly expressing cell lines contained claudin-1 in the nuclear fraction whereas those with no nuclear claudin-1 did not.⁶⁵ The nuclear localization of claudin-1 was also validated by immunostaining and by fractionation followed by Western blots using transformed nasopharyngeal carcinoma cells compared with non-transformed cells, which do not express nuclear claudin-1.⁶⁶ Numerous subsequent studies also showed that claudin-1 localized to the nucleus in melanoma cells, smooth muscle cells from asthmatic subjects, and thyroid carcinoma cells.^67-69 Additionally, claudin-2 localized to the nucleus in human lung adenocarcinoma cells,⁷⁰ claudin-3 localized to the nucleus in breast cancer cell lines,⁷¹ and claudin-4 localized to the nucleus in endometrial cancer cells.⁷² Using genetically modified colon cancer cells, robust expression of nuclear claudin-1 was associated with an increase in invasive capacity, high MMP-9 activity, and a decrease in non-contact mediated cell death (anoikis) rather than affecting cellular proliferation, which is the target of ZO-1 and -2 localized to the nucleus.^63,65,73

The nuclear localization of claudin-1 in colon cancer cells occurred without changes in the molecular weight, suggesting that no cleavage or modification was needed to drive this protein into the nucleus.⁶⁵ In contrast, the prominent nuclear localization of claudin-2 in human lung adenocarcinoma and in cultured lung adenocarcinoma cells was upregulated by dephosphorylation.⁷⁰ Forskolin, via increases in cAMP, was proposed to activate protein phosphatase activity that in turn dephosphorylates claudin-2 to facilitate its nuclear localization.⁷⁰ Ultimately, claudin-2 retained ZONAB and cyclin D1 in the nucleus to enhance cell proliferation.⁷⁰ In melanoma cells, claudin-1 showed both cytoplasmic and weak nuclear localization in isolated cell fractions and by immunostaining.⁶⁷ When claudin-1 was tagged with a strong nuclear localization signal (NLS), the cytoplasmic fraction disappeared in exchange for complete nuclear localization.⁶⁷ This effect was activated by protein kinase A (PKA) and blocked by phosphorylation site mutations, suggesting that claudin-1 also utilizes phosphorylation to regulate its nuclear localization. Different cell types also localize claudin-1 to the nucleus but rather by protein kinase C (PKC) activation.⁶⁹

ZO proteins were shown to have 2 NLS's, one in the PDZ domain and the other in the GK domain, with robust nuclear localization under sparse plating conditions but little localization in confluent monolayers that are contact inhibited.^62,74 Although early work demonstrated that claudins -1 and -2 can gain access to the nucleus, the investigators were unable to show that either claudin had a putative NLS sequence.^65,70 Rather, it was suggested that without an NLS sequence, claudins may utilize their PDZ domain or other mechanism for transport into the nucleus.⁶⁵

While claudin-2 does not have a strong NLS sequence, the cNLS mapper program (see Table 2 for the http:// site) identified putative NLS sequences in many of the 27 known claudin molecules from mouse that could facilitate transport to the nucleus (Table 2). The highest cNLS score was for claudin-12, followed by claudins 10, 23, 15, and 16 (Table 2). For these claudins, and the remainder of claudin molecules with cNLS scores of 3–5, the putative NLS sequence can be localization to the cytoplasm and/or to the nucleus (Table 2). Although it is not known how claudins enter the nucleus, claudin-12 is highly localized to the nucleus in bladder epithelium from mouse and rat, and in human Caco-2 intestinal cells treated with vitamin D.^75,76 Similarily, claudin-15 localized to the nucleus in Caco-2 cells.⁷⁶ Furthermore, claudin-23 is highly expressed in the nucleus of pancreatic cancer cells when compared with its tight junction localization.⁷⁷

Table 2. Putative nuclear localization sequence (NLS) for mouse claudins determined from cNLS Mapper at http://nls-mapper.iab.keio.ac.jp/cgi-bin/NLS_Mapper_form.cgi

Claudin	GenBank or NCBI Reference Sequence	NLS sequence	Position on aa sequence	NLS score^*
Claudin-1	AAC27078.1	RALMVIGILLGLIAIFVSTIGMKCMRC	81	3.5 bipartite
Claudin-3	AAD09756.1	EAQKREMGAGLYVGWAAAALQLLGGALLCCS	153	3.1 bipartite
Claudin-4	AAD09757.1	QKREMGASLYVGWAASGLLLLGGGLLCCS	156	3.4 bipartite
Claudin-6	AAD09759.1	DAQKRELGASLYLGWAASGLLLLGGGLLCC	154	3.5 bipartite
Claudin-7	AAD09760.1	GDDKAKKARIA	109	3.0 monopartite
		RALMVVSLVLGFLAMFVATMGMKCTRC	81	3.3 bipartite
Claudin-8	AAD09761.1	DVALKRELGEALYIGWTTALVLIAGGALFCC	154	3.4 bipartite
Claudin-9	AAD17319.1	EALKRELGASLYLGWAAAALLMLGGGLLCCT	154	3.4 bipartite
Claudin-10	AAD17320.1	IQACRGLMIAAVSLGFFGSIFALFGMKCTKV	76	4.9 bipartite
		DQAKAKIACLAGIVFILSGLCSMTGCSLYANKI	110	3.0 bipartite
Claudin-12	NP_001180590.1	NWRKLRLITFNRNEKNLTIYTGLWVKCARYD	32	5.1 bipartite
Claudin-14	AAD17323.1	RDLQAARALMVISCLLSGMACACAVVGMKCTRC	75	3.6 bipartite
Claudin-15	NP_068365.1	RAGLPYKPSTVVIPRATSDESDISFGKYGKN	194	4.2 bipartite
Claudin-16	NP_444471.1	RNYPYAMRKPYSTAGVSMAKSYKAPRTETAKMYAVD	192	4.1 bipartite
Claudin-17	NP_852467.1	QKRELGGALFLGWATAAVLFIGGGLLCGY	157	3.4 bipartite
Claudin-18 (A1.2)	NP_001181851.1	RALMIVGIVLGVIGILVSIFALKCIRI	80	3.3 bipartite
Claudin-18 (A2.2)	NP_001181852.1	RALMIVGIVLGVIGILVSIFALKCIRI	80	3.3 bipartite
Claudin-19 (Isoform 2)	NP_694745.1	IQSARALMVVAVLLGFVAMVLSVVGMKCTRV	77	3.0 bipartite
Claudin-20 precursor	NP_001095030.1	IQKNQEPWIYPPKQKLSTTWQPKNRRAHN	186	3.9 bipartite
Claudin-22	NP_083659.1	EGLKKRLLTLGGTLLWTSGVMVLVPVSWVAHK	112	3.8 bipartite
Claudin-23	NP_082274.1	PAIKYYSDGQHRPPPTAEHRDTSKLKVGFP	219	4.7 bipartite
Claudin-24 (putative)	NP_001104788.1	EGLKKRLLTLGGTLLWTSGVMVLVPVSWVAHK	112	3.8 bipartite

Note.

* A score of 3–5 indicates that the sequence can be localized to both nucleus and cytoplasm. Monopartite: NLSs contain a single cluster of basic residues and are divided into 2 subclasses; one with at least 4 consecutive basic amino acids (class 1) and the other with 3 basic amino acids, represented by K(K/R)X(K/R) as a putative consensus sequence where X indicates any amino acid (class 2). Bipartite: NLSs contain 2 clusters of basic residues separated by a 10–12 amino acid linker. A putative consensus sequence of the bipartite NLS has been defined as (K/R)(K/R)X10–12(K/R)3/5 where X indicates any amino acid and (K/R)3/5 represents at least 3 of either lysine or arginine out of 5 consecutive amino acids. This information was provided by the URL site, above. Further information about the methods used to derive the sequences and scores can be obtained from that site.

The nuclear localization of claudins suggests that these proteins function to regulate gene transcription but is there any evidence to support this notion? Molecular studies by Dhawan et al,⁶⁵ using cultured colon cancer cell lines, lend considerable insight. For this, the nuclear localization of claudin-1 in genetically manipulated colon cancer cell lines, from highly expressing to virtual knockouts, resulted in 1) changes in luciferase activity of the E-cadherin promoter that were consistent with the level of nuclear claudin-1 expression (inhibition of nuclear claudin-1 expression resulted in the activation of E-cadherin reporter activity) and 2) concomitant changes in β-catenin/Tcf/Lef and Wnt/β-catenin/Tcf/Lef signaling that affected the expression of downstream target genes such as c-Myc.⁶⁵ Overall, these studies demonstrate a direct transcriptional role for nuclear claudin-1 in E-cadherin expression that affected indirectly the β-catenin/Tcf/Lef signaling pathway in cultured colon cancer cells.⁶⁵

Although it seems puzzling that a membrane-spanning protein as large as claudin molecules (20–27 kDa) would transport and enter the nucleus, there is emerging data to support that similarly sized tetraspanin molecules do so and also regulate gene transcription. It was recently shown that TM4SF3 (26 kDa), also known as tetraspanin-8, enters the nucleus of prostate cancer cells.⁷⁸ The efficiency of nuclear accumulation is dependent on a direct interaction with the androgen receptor, which provides stability for TM4SF3.⁷⁸ Once in the nucleus, TM4SF3 contributes to androgen-dependent gene expression, particularly to the expression of genes regulating cellular proliferation.⁷⁸ CD9 (25 kDa), also known as tetraspanin-29, was shown to enter the nucleus in breast cancer cells.⁷⁹ CD9 associated with IgFS8 in the cytoplasm and IgFS8 and CEP97 in the nucleus, suggesting that CD9 is part of a larger complex that, in some way, translocates from the cell membrane to the nucleus to regulate gene transcription.⁷⁹ Recent studies demonstrated that CD9 transits to and enters the nucleus using a novel extracellular vesicle-derived system that delivers cargo to the nucleus via Rab7-positive late endosomal vesicle fusion with the nuclear envelope.⁸⁰ E-cadherin, a type-1 membrane spanning protein, also has a nuclear localization pattern. To gain access to the nucleus, the large E-cadherin molecule is cleaved by metaloproteinases and by γ-secretase into E-cadherin C-terminal fragments, which along with a chaperone protein (p120) translocates to the nucleus and regulates gene transcription.^81,82 Results from Dhawan et al⁶⁵ argue against claudins entering the nucleus as fragments because the band for claudin-1 in Western blots was the same size in membrane-derived, cytoplasmic, and nuclear fractions of colonic cancer cells.

Overall, many claudins have a putative NLS, or multiple NLS's, which map to the intermembrane domain alone or to the extracellular loop 2/intermembrane domain area of the claudin structure⁸³ that would facilitate transport into the nucleus. Although constitutive and disease-related shuttling of claudins into the nucleus occurs, little is known about conditions that facilitate this event. Although the existing data support that nuclear claudin molecules interact with transcriptional regulators to impact gene expression and the regulation of cell attachment, cell proliferation/cell death, and cell proliferation effectors, further data are needed to identify putative binding partners in the cytoplasm and in the nucleus.

Conclusions

There is a wealth of supporting data in the literature to conclude that claudin molecules are purposely localized to sites outside of the apical tight junction complex and as such, serve important roles in both mucosal homeostasis and in disease pathogenesis. In particular, most if not all of the claudin family members localize to numerous other sites, being renegades by not fitting into a single functional paradigm at the tight junction.

To move this field ahead, it is possible that new tools may be needed to properly determine whether or not the basolateral pool of claudins is a reservoir of extra material that can be recruited to the tight junction when needed. If it is, we also need to understand what regulates this function. Furthermore, it is essential to understand whether or not basolateral claudins are actually expressed within the tight junction complex, or is the proximity so close that our current instrumentation cannot resolve them spatially. If solely along the basolateral membrane, do basolateral claudins form a secondary permeability barrier that acts in series to the tight junction barrier to lend additional transit time for the molecules crossing into the basolateral compartment? Do basolateral claudins, via their PDZ domain, nucleate a yet undefined macromolecular complex that organizes signaling or other functions to the basolateral membrane and do they also use this feature of their structure to gain access to the nucleus? What is the role of cytoplasmic and nuclear claudins and how are these claudins regulated? Do nuclear claudins directly regulate transcription by forming DNA binding complexes within the nucleus or indirectly regulate transcription, via signaling pathways from the cytoplasm? Similarly, do the nuclear claudins work in concert to promote or inhibit the action of classic transcription factors? These questions and numerous others pose an exciting new era in barrier biology that will take us well into the future.

Abbreviations

CLDN	=	claudin
ECM	=	extracellular matrix
EpCAM	=	epithelial cell adhesion molecule
FAK	=	focal adhesion kinase
GI	=	gastrointestinal
JAM	=	junctional adhesion molecule
MAPK	=	MAP kinase
MMP	=	matrix metalloproteinase
NLS	=	nuclear localization signal (or sequence)
PKA	=	protein kinase A
PCK	=	protein kinase C
TGN	=	trans-Golgi network
ZO	=	zonula occludins.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Funding

This work was supported by the Harvard Catalyst | The Harvard Clinical and Translational Science Center, National Center for Research Resources and the National Center for Advancing Translational Sciences, National Institutes of Health (NIH) Award 1UL1 TR001102–01 and financial contributions from Harvard University and its affiliated academic health care centers, Department of Surgery Bridge funds, Special Fund for Gastric Cancer Research, and NIH grant P30 DK034854.