*Review* **A Novel Claudinopathy Based on Claudin-10 Mutations**

#### **Susanne Milatz \***

Institute of Physiology, Kiel University, Christian-Albrechts-Platz 4, 24118 Kiel, Germany

Received: 15 October 2019; Accepted: 27 October 2019; Published: 30 October 2019

**Abstract:** Claudins are key components of the tight junction, sealing the paracellular cleft or composing size-, charge- and water-selective paracellular channels. Claudin-10 occurs in two major isoforms, claudin-10a and claudin-10b, which constitute paracellular anion or cation channels, respectively. For several years after the discovery of claudin-10, its functional relevance in men has remained elusive. Within the past two years, several studies appeared, describing patients with different pathogenic variants of the *CLDN10* gene. Patients presented with dysfunction of kidney, exocrine glands and skin. This review summarizes and compares the recently published studies reporting on a novel autosomal-recessive disorder based on claudin-10 mutations.

**Keywords:** tight junction; paracellular permeability; paracellular sodium transport; thick ascending limb; nephropathy; HELIX syndrome; hypokalemia; hypermagnesemia; anhidrosis; gland dysfunction

#### **1. Introduction**

#### *1.1. Claudin-10*

The protein family of claudins is a key component of the tight junction (TJ). Claudins comprise four transmembrane segments (TM1-4), two extracellular segments (ECS1 and 2) and intracellular N- and C-termini. Embedded in the plasma membranes of adjacent cells, they interact with each other within the same plasma membrane but also across the paracellular cleft, with claudins of the neighboring cell (cis- or trans-interaction, respectively). By this means, they form a complex strand meshwork and determine tightness and selectivity of the bicellular TJ. Whereas most claudins exhibit a mainly sealing function, some claudins form size-, charge- and water-selective channels through the TJ [1–18].

The claudin family encompasses at least 24 members in mammals. The human gene encoding claudin-10 (*CLDN10*) contains six exons and gives rise to two major isoforms: claudin-10a and -10b. According to their usage of either exon 1a or 1b, they differ in their TM1 and ECS1 (Figure 1). As ECS1 acts as main determinant of charge selectivity, claudin-10a and -10b strands exhibit contrarian permeability properties.

**Figure 1.** Predicted topology of claudin-10. The two major isoforms claudin-10a and claudin-10b differ in their first transmembrane segment and most of the first extracellular segment (ECS1), both shown in light grey. The remaining protein sequence is identical (shown in dark grey). The mutations discovered to date (red) comprise single amino acid substitutions or large deletions (M1T, ΔE4) and affect either both claudin-10a and claudin-10b or only claudin-10b. The existing claudin-10a variants with respective residue numbering are depicted in parentheses.

Due to claudin-10a's equipment with seven positive and only one negative amino acid in ECS1, it is predestined to form a paracellular anion channel [10,12]. Expression of human claudin-10a in the poorly ion permeable cell line MDCK C7 resulted in a decrease in transepithelial resistance (TER) without alteration in preference for Na<sup>+</sup> or Cl– [12]. Moreover, claudin-10a expression increased the relative NO3<sup>−</sup> permeability but decreased the permeability to the anion pyruvate, suggesting that claudin-10a modifies the anion preference of the paracellular pathway [12].

ECS1 of claudin-10b comprises four positive and five negative amino acids. In most cell culture models, heterologous expression of human or murine claudin-10b led to a marked decrease in TER that was based on an increase in Na<sup>+</sup> permeability over Cl– permeability (PNa+/PCl–). Further studies revealed a relative strong permeability to all monovalent cations with preference for Na+, a lesser permeability to divalent cations and impermeability to larger molecules (4 kDa dextran) or water of the claudin-10b-based paracellular channel [10,12,13,19].

Claudin-10a and -10b do not only differ significantly in their function but also with respect to their expression in the body. Whereas claudin-10a appears to be restricted to the kidney, claudin-10b has been detected in many tissues, including kidney, skin, salivary glands, sweat glands, brain, lung and pancreas [10,12,20]. Along the kidney tubule, claudin-10a is expressed in the proximal tubule to the S3 segment whereas the main expression site of claudin-10b is the thick ascending limb of Henle's loop (TAL). Claudin-10b is found along the whole medullary–cortical axis of TAL from inner stripe of outer medulla (ISOM) to outer stripe of outer medulla (OSOM) to the renal cortex. In ISOM TAL, to current knowledge, solely claudin-10b constitutes the bicellular TJ, where it facilitates Na<sup>+</sup> reabsorption [21,22]. Towards OSOM and cortex, a TAL mosaic claudin expression is found. Claudin-10b equips part of the TJs, whereas the remaining TJs contain a complex of claudin-3, -16, -19 and to a smaller extent claudin-14 [21–23]. This complex is involved in the reabsorption of divalent cations such as Ca2<sup>+</sup> and Mg2<sup>+</sup> in the TAL. The thin limb of Henle also incorporates claudin-10, as yet the identity of the present isoform is unknown [24].

An important insight into the physiological role of claudin-10b was provided by the mouse model generated by Breiderhoff et al., lacking claudin-10 in the entire loop of Henle. These mice featured a strongly reduced paracellular Na<sup>+</sup> selectivity in the TAL that led to a urinary concentration defect and was accompanied by hypermagnesemia, polyuria, polydipsia, elevated plasma urea levels and compensatorily increased K<sup>+</sup> and H<sup>+</sup> secretion [25,26]. This misbalanced TAL electrolyte handling was accompanied by a severe medullary nephrocalcinosis in these animals.

#### *1.2. Claudinopathies*

So far, a number of human hereditary diseases based on defects in claudin-1, -14, -16 and -19 have been reported [27–32]. However, for 18 years after the discovery of claudin-10, its functional relevance in men has remained unclear. On the one hand, defects in the *CLDN10* gene are rare and clinical manifestations occur mainly in patients with biallelic defects (autosomal recessive disorder). On the other hand, some patients presented with symptoms many years ago but were originally misdiagnosed with Bartter syndrome or Gitelman syndrome. Both diseases are characterized by a salt-losing nephropathy with an imbalance in Ca2<sup>+</sup> and Mg2<sup>+</sup> homeostasis. In Bartter syndrome, the transcellular NaCl reabsorption in the TAL is disrupted, due to mutations in the Na-K-Cl cotransporter 2 (NKCC2), the renal outer medullary potassium channel (ROMK1) or the Cl– channel Kb (ClC-Kb) [33–36]. Gitelman syndrome is caused by mutations affecting the Na+-Cl– cotransporter (NCC) in the distal convoluted tubule [37]. Nowadays, whole-exome sequencing is available at lower cost and increasingly used to identify the cause of rare mendelian disorders. In 2017, three studies reported on patients with different pathogenic variants of *CLDN10* [38–40]. Hadj-Rabia et al. coined a novel disease syndrome, summarizing the clinical manifestations of their patients (HELIX for hypohidrosis, electrolyte imbalance, lacrimal gland dysfunction, ichthyosis, xerostomia) [39]. This review aims to summarize and compare the data of Bongers et al., Hadj-Rabia et al., Klar et al. and a case report by Meyers et al., all describing a novel claudinopathy based on *CLDN10* mutations [38–41].

#### **2. Clinical Manifestations**

To date, a total of 22 patients carrying homozygous or compound heterozygous *CLDN10* mutations have been reported, their ages ranging from 4 to 53 years. Patients were mostly born in consanguineous families and first symptoms often occurred in early childhood, sometimes directly after birth. Patients presented with first symptoms as anhidrosis, xerostomia, alacrima, muscle cramps, falls, or chest pain. Table 1 provides a summary of all patient groups and their clinical manifestations.


**Table 1.** Summary of all patient groups, their *CLDN10* variants and clinical manifestations.


**Table 1.** *Cont.*

\* with the exception of normal deposition in eccrine sweat glands. \*\* considering the patient's age. \*\*\* compared to heterozygous family members. n.r. not reported

#### *2.1. Hypohidrosis, Xerostomia and Alacrima*

Hypohidrosis with intolerance to heat was frequently reported as one of the first symptoms observed in early childhood. Apparently, all known patients suffer from hypohidrosis, including the two patients described by Bongers et al., who did not complain about reduced sweating at the outset but confirmed hypohidrosis subsequently [38], (personal communication with Tom Nijenhuis, Radboud University Medical Center, Nijmegen). Klar et al. evaluated heat intolerance in two patients by exposure to heat for 20 min, which resulted in a rapidly increased body temperature from 37 ◦C to 39.6 ◦C and an increase in heart rate [40]. Generalized hypohidrosis was verified using starch-iod test applied on different body parts, corroborating a severe dysfunction of eccrine sweat glands.

Likewise, xerostomia due to reduced saliva production is apparently a typical symptom of *CLDN10* defects as it has been documented in all known patients including the patients examined by Bongers et al. [39–41]; (personal communication with Tom Nijenhuis). Hadj-Rabia et al. analyzed xerostomia by saliva secretion rate measurements in three adult patients [39]. As a result, the flow of fluid was reduced by 98% in patients compared to healthy controls. Moreover, the fluid/mucus ratio of saliva was dramatically decreased in patients with *CLDN10* variants. Hadj-Rabia also documented a poor dental condition with severe enamel wear and generalized gingival inflammation of their patients [39]. This is attributed to aptyalism but might also be a consequence of disturbed enamel mineralization (amelogenesis imperfecta) as claudin-10 expression was found in ameloblasts of mice [42].

Alacrima (the inability to produce tears) was described in the majority of patients and was confirmed using Schirmer's test by Hadj-Rabia et al. [39–41].

#### *2.2. Dermatological Manifestations in Addition to Hypohidrosis*

The occurrence of ichthyosis and other dermatological manifestations among patients was rather inconsistent. Hadj-Rabia described mild forms of ichthyosis with a thickened stratum corneum, palmar hyperlinearity and plantar keratoderma in two unrelated families with different *CLDN10* variants. Histological analysis of skin biopsies revealed slight epidermal hyperplasia and an abnormally high number of dilated eccrine sweat glands. The patients examined by Bongers et al. reported dry skin in retrospect, and dermatological consultation showed palmar hyperlinearity and plantar hyperkeratosis (personal communication with Tom Nijenhuis). In contrast, patients examined by Klar et al. and Meyers et al. showed no dermatological manifestations apart from hypohidrosis [40,41]. Morphology and number of eccrine sweat glands appeared normal in patients examined by Klar et al. [40].

#### *2.3. Kidney Dysfunction*

A number of features revealing a renal dysfunction in patients has been reported. All patients showed a disturbance in electrolyte balance, becoming manifest in hypermagnesemia, hypokalemia and hypocalciuria. Hypermagnesemia was present in the majority of patients, most pronounced in childhood and decreasing with age (Figure 2A). In contrast, hypokalemia was most severe in adults with the exception of the patient group examined by Klar et al. [40],(Figure 2B). These patients had serum K<sup>+</sup> levels in the upper normal range or higher in adulthood. The reason for that discrepancy remains unclear. Admittedly, more longitudinal intraindividual data over a longer time period are required for confirmation of the apparent age dependencies depicted in Figure 2. Bongers et al. reported that hypokalemia was accompanied by metabolic alkalosis in their patients [38].

**Figure 2.** Available plasma/serum values of patients plotted against their age. (**A**) Hypermagnesemia is most pronounced in children and decreases with age. (**B**) Plasma/serum potassium levels decline with age, revealing a marked hypokalemia in adolescence. The only exception is the patient group examined by Klar et al [40]. Values were partly obtained by conversion into mmol/L. The publication by Bongers et al. provided data points of the same patients at different ages [38].

Of note, hypocalciuria appears to be a common manifestation as it was distinct in all tested patients and at all ages with plasma/serum Ca2<sup>+</sup> in the normal or upper normal range. Despite the strong incidence of hypermagnesemia and hypocalciuria, most patients showed no decrease in urinary Mg2<sup>+</sup> or an elevation of plasma/serum Ca2+. The reason for this is not clear. However, it appears that renal Ca2<sup>+</sup> and Mg2<sup>+</sup> reabsorption and handling are tightly but differentially regulated.

In some patients, plasma aldosterone levels were determined and revealed hyperaldosteronism [39, 41]. The underlying cause of the electrolyte disorder and hyperaldosteronism is a NaCl wasting in the TAL, the main expression site of claudin-10b. Plasma/serum Na<sup>+</sup> appeared mostly normal if tested, presumably due to compensation in the more distal parts of the nephron. However, slight hypochloremia was present in few patients [39–41].

Bongers et al. tested the urine concentrating ability of one patient [38]. As a result, the patient showed a reduced response to thirsting and the application of the synthetic vasopressin analogue ddAVP. Verification of an adequate aquaporin-2 response in the collecting duct pointed to a dysfunction of NaCl reabsorption in the TAL as cause for an insufficient build-up of interstitial tonicity. In line with the reduced urine concentrating ability, this patient suffered from polyuria. Overall, the occurrence of polyuria and polydipsia among patients was rather inconsistent and might also be partially attributed to xerostomia.

The estimated glomerular filtration rate (eGFR) as indicator for kidney function was determined in most patients. It ranged from low to normal and decreased with time in some patients, indicating a progressive renal insufficiency. With one exception, kidneys analyzed by computer tomography scans were normal in form and size and, in contrast to the mouse model, nephrocalcinosis was not detected in any patient [38–41]. However, several patients of Klar et al. suffered from recurrent kidney pain due to nephrolithiasis [40]. If determined, the patient's blood pressure ranged from low to normal. Some patients complained of atypical chest pain, heart palpitations, collapse, falls, or muscle cramps. Klar et al. additionally analyzed lung and pancreas function without abnormal findings [40].

#### **3. Protein Variants and Their Functional Analyses**

Figure 1 displays all naturally occurring claudin-10 variants reported so far. Four mutations apply to both claudin-10a and claudin-10b proteins, whereas three defects affect only claudin-10b. Markedly, there are no obvious differences between the six patients with both defective isoforms and the 16 patients with defective claudin-10b and unaffected or only heterozygously affected claudin-10a, at least none that could be easily attributed to that special factor (Table 1). Of note, to date no patients with exclusive claudin-10a variants (without affection of claudin-10b) were described, also pointing to a minor role of claudin-10a defects in men.

Because of the predominant importance of claudin-10b defects in patients, most in vitro studies were carried out using the mutants of this isoform. Several analyses addressed localization, strand formation and channel function of mutated claudin-10b vs. wildtype protein. In the following, the denomination of claudin-10 variants refers to the alteration in claudin-10b protein, i.e., the amino acid exchange or deletion.

#### *3.1. Membrane Localization*

Correct trafficking to the membrane of a claudin variant is a fundamental prerequisite for a physiological function within the TJ. Similar to the wild type protein, the claudin-10b N48K mutant analyzed by Klar et al. resided in the cell membrane when heterologously expressed in the epithelial cell line MDCK C7. However, the subcellular distribution of N48K suggested an increased intracellular accumulation [40]. A very similar distribution in MDCK C7 cells was observed for the variants D73N and P149R investigated by Bongers et al. [38]. In sharp contrast, their third variant ΔE4 was very weakly expressed and did not localize to the cell membrane. This variant lacks TM4 which inevitably leads to exclusion from the plasma membrane. Likewise, the variant S131L with a missense mutation in TM3 analyzed by Hadj-Rabia et al. showed no detectable presence at the plasma membrane when expressed in a mouse TAL cell culture [39]. Contradictorily, the claudin-10 variant M1T probably lacking part of TM1 revealed normal deposition in eccrine sweat glands of a skin biopsy but was not localized in the TAL of a kidney biopsy [39]. Klar et al. also performed immunostainings of sweat glands and observed a normal localization of N48K in membranes facing the lumen but also an abnormally strong intracellular distribution without accumulation in canaliculi, pointing to a reduced delivery to the TJs [40].

Overall, the analyzed claudin-10b mutations resulted either in complete absence from the plasma membrane (especially those affecting one of the transmembrane segments) or, on the other hand, can basically insert into the plasma membrane, although an increased distribution outside the TJ suggests a reduced function.

#### *3.2. TJ Strand Formation*

Strand formation is the consequence of cis- and trans-interaction between claudins. In case of claudin-10b and its mutants, interaction with itself (homophilic interaction) is of particular importance as claudin-10b is not capable of interaction with any other claudin in the TAL [22]. Trans-interaction (with claudins in the opposing plasma membrane) can be detected by heterologous expression of the appropriate claudin in TJ-free HEK 293 cells and subsequent microscopic analysis of so-called contact enrichment. If the claudin is capable of autonomous strand formation, it enriches at cell–cell-contacts of two transfected cells compared to the remaining cell membrane. Bongers et al. reported a basal capability of trans-interaction and TJ formation for variants D73N and P149R but not for the truncated variant ΔE4 [38]. Klar et al. did not detect a significant contact enrichment of their mutant N48K, tantamount to a perturbed homophilic trans-interaction [40]. Moreover, homophilic cis-interaction was analyzed by Förster resonance energy transfer (FRET) and revealed an exaggerated oligomerization of N48K proteins. Ultrastructural analysis using freeze fracture electron microscopy showed that N48K formed few particle-typed TJ strands with less compact meshworks, compared to wildtype claudin-10b [40].

Overall, the few claudin-10b variants analyzed with respect to interaction properties mostly showed a fundamental capability of homophilic interaction and strand formation. Nonetheless, quality and/or quantity of interaction and strand assembly were impaired. As anticipated, the lack of a transmembrane segment resulted in a complete loss of function.

#### *3.3. Channel Function*

Analyses of claudin-10b mutant channel properties by means of electrophysiological measurements are available only for the N48K variant, as yet. Klar et al. stably transfected MDCK C7 cells and compared claudin-10b N48K-with wildtype-expressing cells [40]. The cation selectivity (PNa/PCl) of N48K-expressing cells was first similar to that of claudin-10b wildtype-transfected cells but progressively decreased with passaging, more and more resembling that for clones with a weak expression of claudin-10b wildtype. Moreover, N48K-expressing cells showed a higher TER and altered relative permeabilities to other monovalent cations, compared to claudin-10 wildtype-expressing cells. Together, the results suggest that the channels formed by N48K have a subnormal Na<sup>+</sup> permeability and that the overall number of channels is markedly reduced [40].

In an attempt to mimic sweat secretion, Klar et al. used a 3D cell culture model by growing MDCK C7 cells in Matrigel. Epithelial cells formed three-dimensional cysts with the apical side towards the lumen. Cysts formed by wildtype-claudin-10b-expressing cells increased their lumen by transcellular Cl<sup>−</sup> secretion, followed by claudin-10b-mediated Na<sup>+</sup> permeation and subsequent water transport. N48K-expressing cysts, in contrast, showed considerably less lumen expansion, indicating a reduced overall Na<sup>+</sup> conductance. In line with electrophysiological studies, this is probably caused by a combination of a reduction in single channel Na<sup>+</sup> conductance and a reduction in the overall number of channels [40].

#### *3.4. Protein Structure*

In order to determine the cause of the impairments brought about by single amino acid substitutions in claudin-10b, Klar et al. and Hadj-Rabia et al. provided 3D homology models of N48K or S131L, respectively, based on the crystal structure of murine claudin-15 [39,40,43]. The N48K mutation, localized in ECS1, appears to disrupt an intramolecular bridging between different backbones in ECS1 and membrane–ECS1-transition. Moreover, a potential electrostatic interaction could be disturbed by the replacement of an uncharged residue by a positively charged one. These alterations are presumed to perturb protein interaction and function indirectly [40]. The S131L mutation analyzed by Hadj-Rabia et al. affects TM3 and is suggested to clash sterically with several of the surrounding residues, especially of TM1 and -2 and by this perturbing the compactness of the helical bundle [39]. Furthermore, an intrahelical stabilizing hydrogen bond is lost by the amino acid substitution. As a consequence of the helical bundle destabilization, the S131L variant fails to insert into the cell membrane and is retained in the cytosol.

The impact of the amino acid substitutions D73N, R80G and P149R on protein folding or pore formation remain temporarily unsolved, also because functional data on the particular variants are scarce or absent. In general, substitution of a charged residue by an uncharged one in ECS1 (D73N, R80G) is considered critical for ion pore formation. On the other hand, neither D73 nor R80 have been shown to be important for charge selectivity of pore-forming claudins (for review see [44]). However, R80 is localized at the predicted transition between ECS1 and TM2 and might be crucial for the formation of the helical bundle. P149 in ECS2 is conserved in the majority of claudins. It is suggested to stabilize ECS2 conformation and to play a role in correct TJ strand arrangement [45].

#### *3.5. Short Summary of Functional Analyses*

Taken together, claudin-10b variants with truncations in one of the transmembrane segments necessarily show a complete loss of function as they are not inserted into the plasma membrane whereas variants with point mutations in one of the loops often keep a certain residual function. Nevertheless, localization, interaction and channel function can be impaired to a certain extent. Point mutations in transmembrane segments still have a high probability to severely affect membrane localization, probably depending on the substituting residue. The comparatively poor viability of cells expressing claudin-10b mutant proteins compared to wildtype-expressing cells and the increased degradation of mutant proteins due to suboptimal folding and localization as observed for several variants appear to be a common incidence and presumably contribute pivotally to the harmfulness of mutations [38–40].

Of course, the question arises to which extent site and type of claudin-10b mutation determine the clinical outcome. The most prominent differences between patient groups with particular *CLDN10* variants concern plasma/serum potassium values and the dermatological phenotype. However, with the actual number of patients and the information available it is difficult to assess a possible correlation and future studies will probably help to clarify that issue.

#### **4. Mechanisms of Disease**

The naturally occurring claudin-10 mutations discovered during the last years result in a complete or partial loss of function of the claudin-10b isoform and subsequently in a reduced or absent paracellular Na<sup>+</sup> permeation in the TAL as well as in sweat, salivary and lacrimal glands. Figures 3 and 4 illustrate the mechanistic principles, particularly the role of claudn-10b in epithelial transport, underlying a correct organ function.

#### *4.1. Kidney*

As mentioned above, the exact role of claudin-10a in the proximal tubule is not understood, as yet. Based on cell culture experiments, a function as paracellular anion channel is assumed (Figure 3A). However, the data presented in the reviewed studies indicate a rather minor role in the pathogenesis of claudin-10 defects. It appears likely that possible impairments affecting the proximal tubule can be distally compensated to a certain degree as it was shown for the loss of claudin-2 in a mouse model [46]. After all, future studies will have to clarify the physiological relevance of claudin-10a.

In the TAL, half of the Na<sup>+</sup> is reabsorbed paracellularly via the claudin-10b paracellular channel as a consequence of transcellular net uptake of Cl– involving NKCC2 (Figure 3B). In TAL of the ISOM, claudin-10b dominates the TJ, although claudin-3, -16 and -19 are expressed intracellularly. Towards OSOM and cortex, an additional epithelial cell type occurs, expressing claudin-3, -16 and -19 but no claudin-10b. These cells form TJ complexes that are spatially separated from claudin-10b TJs and are involved in the reabsorption of Mg2<sup>+</sup> and Ca2<sup>+</sup>. The residual TJs assembled exclusively by claudin-10b enable a backflux of Na<sup>+</sup> into the lumen, due to its concentration gradient, thus adding to the lumen-positive potential and supporting the paracellular reabsorption of Mg2<sup>+</sup> and Ca2<sup>+</sup> [21,22,47].

An impaired claudin-10b function would reduce the Na<sup>+</sup> reabsorption in the TAL by the paracellular portion and lead to a compensatorily increased electrogenic Na<sup>+</sup> reabsorption via the epithelial sodium channel (ENaC) in the more distal nephron segments. This, in turn, would promote K<sup>+</sup> and H<sup>+</sup> loss. Whereas hyperaldosteronism, hypokalemia and metabolic alkalosis in patients with pathogenic claudin-10b variants are the consequence of compensatory mechanisms in the distal nephron, hypermagnesemia and hypocalciuria can be attributed to an exaggerated paracellular reabsorption of Mg2<sup>+</sup> and Ca2<sup>+</sup> in the TAL. In the mouse model described by Breiderhoff et al., the lack of claudin-10b in the TAL results in an increased distribution of the TJ complex containing claudin-16 and -19 over all TJs, including those in the ISOM, where normally only claudin-10b is constituting the TJ [22,25,26,48]. If a similar process takes place in men with *CLDN10* variants that lead to a reduced claudin-10b insertion into TAL TJs, it would explain the hypermagnesemia, in line with the mouse model. Apparently, in patients, the reabsorption of Mg2<sup>+</sup> and Ca2<sup>+</sup> is shifted from cortex/OSOM TAL to ISOM TAL and the lack of claudin-10b-based Na<sup>+</sup> backflux as additional driving force in cortex/OSOM TAL is thereby overcompensated.

**Figure 3.** Mechanisms of claudin-10 function in the kidney. (**A**) Claudin-10a is expressed in the renal proximal tubule to the S3 segment (shown in blue) where it forms a paracellular anion pathway. Patients with pathogenic variants of both claudin-10a and claudin-10b appear to have no clinical manifestations different from patients with only claudin-10b variants, suggesting a minor role of an impaired claudin-10a function and/or compensation in more distal nephron segments. (**B**) Claudin-10b is mainly localized in the thick ascending limb of Henle's loop (TAL, shown in red) where it facilitates the paracellular reabsorption of Na+. Na<sup>+</sup>, 2 Cl– and K<sup>+</sup> enter the epithelial cell by the secondary active NKCC2 transporter. Whereas Na<sup>+</sup> and Cl– leave the cell basolaterally, K<sup>+</sup> is recirculated via the ROMK1 channel. This results in a net reabsorption of Cl– and causes a lumen-positive transepithelial potential, which in turn drives paracellular Na<sup>+</sup> reabsorption via claudin-10b. Towards the distal TAL segment the expression of claudin-16 and -19 is involved in the paracellular reabsorption of Mg2<sup>+</sup> and Ca2+. Here, Na<sup>+</sup> can backflux into the lumen along its concentration gradient, thus adding to the lumen-positive potential as a driving force for Mg2<sup>+</sup>- and Ca2<sup>+</sup> reabsorption. An impairment of claudin-10b function would result in a NaCl wasting in the TAL and an expansion of claudin-16 and -19 over all TJs, causing hypermagnesemia and hypocalciuria. (**C**) Na<sup>+</sup> wasting in the TAL is compensated in the more distal nephron involving electrogenic transport via ENaC, promoting secretion of K<sup>+</sup> and causing hypokalemia.

However, the reasons for the magnitude of hypokalemia and hypermagnesemia depending on the patients' age remain unsolved.

#### *4.2. Glands*

In glands, a concerted secretion of ions and water across the glandular epithelium is required in order to produce sweat, saliva or lacrimal fluid, respectively (Figure 4). Molecular mechanisms in the secretory coil of sweat glands as well as in salivary and lacrimal acinar cells are partially understood and share certain mechanistic similarities [49–56]. Stimulation of G-protein-coupled muscarinergic receptors by acetylcholine results in an increased intracellular concentration of free Ca2+, which activates apical Cl<sup>−</sup> channels and basolateral K<sup>+</sup> channels. Subsequent activation of NKCC1 in the basolateral membrane leads to an influx of Na+, K<sup>+</sup> and Cl– into secretory cells. Na<sup>+</sup> and K<sup>+</sup> circulate across the basolateral membrane via NKCC1, the Na+/K+-ATPase and/or K<sup>+</sup> channels, respectively. The transcellular net flux of Cl− via apical channels and basolateral NKCC1 generates a lumen-negative transepithelial potential and drives the paracellular secretion of Na<sup>+</sup> via claudin-10b channels. This is followed by transcellular water secretion involving aquaporin-5 water channels. Although an additional paracellular water secretion is frequently suggested, the strict water impermeability of claudin-10b-based TJs [13] and the implausibility of claudin-10b to form heterogeneous TJs with other claudins [22] rather argue against it.

An impairment of claudin-10b distribution or function as reported by Klar et al. would result in a complete abrogation of Na<sup>+</sup> and subsequent water secretion in sweat, salivary and lacrimal glands and become manifest in hypohidrosis, aptyalism and alacrima as observed in patients [40].

**Figure 4.** Mechanisms of claudin-10b function in the secretory portion of sweat glands and the acinar cells of salivary and lacrimal glands. During secretion, apical Cl– channels and basolateral K<sup>+</sup> channels are activated. Na<sup>+</sup>, 2 Cl– and K<sup>+</sup> enter the cell via the basolaterally expressed NKCC1 transporter. Cl– is secreted into the lumen whereas Na<sup>+</sup> and K<sup>+</sup> leave the cell basolaterally. The secretion of Cl– drives Na<sup>+</sup> transport via the paracellular claudin-10b channel and transcellular water transport via aquaporins. An abrogation of the claudin-10b-mediated Na<sup>+</sup> transport would result in a complete dysfunction of sweat, saliva and tear secretion.

#### *4.3. Skin*

At present it remains unsolved whether dermatological manifestations such as ichthyosis and dry skin are a mere consequence of sweat gland dysfunction or may also be augmented by defective claudin-10 proteins in the epidermis. The mechanical barrier function of the skin is maintained on the one hand by the stratum corneum containing dead, cornified cells (corneocytes) and by the TJs on the other hand (for review see [57]). Functional TJs are localized in the stratum granulosum where they form a liquid–liquid interface barrier. The role of claudins in epidermal barrier function is illustrated by the phenotype of patients with pathogenic variants of the *CLDN1* gene. Patients lacking functional claudin-1 suffer from NISCH syndrome (neonatal ichthyosis–sclerosing cholangitis). The thick, scaly skin of ichthyosis is assumed to result from the skin compensating for barrier dysfunction [58]. On the other hand, claudins are expressed not only in the stratum granulosum but in all living layers of the epidermis and are probably involved in the differentiation of stratum granulosum cells to corneocytes.

Claudin-10 mRNA was detected in human skin biopsies and was found to be downregulated in patients with psoriasis vulgaris [59–61]. Immunohistochemical staining of rodent skin revealed an intracellular localization in the stratum basale [20] or in the stratum corneum [62]. The finding that claudin-10 is expressed beyond the stratum granulosum points to a possible function aside from its role as an ion pathway in the epidermis. However, its particular function in the skin has yet to be unraveled.

#### **5. Summary**

The four studies of Bongers et al., Klar et al., Hadj-Rabia et al. and Meyers et al. describe a novel claudinopathy that is based on mutations in the *CLDN10* gene and is characterized by an impaired function of mainly claudin-10b. Patients show a salt-losing nephropathy without hypercalciuria (as in Bartter syndrome) or hypomagnesemia (as in Gitelman syndrome). Main symptoms probably occurring in the majority of present and future patients are:


The severity of clinical manifestations may depend on site and type of mutation. Patients expressing a claudin-10b variant with a defect in one of the transmembrane regions, especially a truncation, are at high risk to develop the full severity of the phenotype. In this respect, genotyping of future patients may further help to correlate mutation type and course of disease. Further functional analysis of found *CLDN10* variants may provide a more detailed insight into the protein function. The reviewed studies demonstrate the significance of proper claudin-10b function for kidney, skin and gland physiology.

**Funding:** This work received no external funding.

**Acknowledgments:** I thank Nina Himmerkus (Institute of Physiology, Kiel University, Kiel, Germany) and Tom Nijenhuis (Radboud University Medical Center, Nijmegen, The Netherlands) for reading the manuscript draft and for helpful discussions and suggestions.

**Conflicts of Interest:** The author declares no conflict of interest.

#### **References**


© 2019 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Claudin-2: Roles beyond Permeability Functions**

#### **Shruthi Venugopal, Shaista Anwer and Katalin Szászi \***

Keenan Research Centre for Biomedical Science of the St. Michael's Hospital and Department of Surgery, University of Toronto, Toronto, ON M5B 1W8, Canada; shruti.venugopal@gmail.com (S.V.); shaista.anwer@mail.utoronto.ca (S.A.)

**\*** Correspondence: katalin.szaszi@unityhealth.to; Tel.: +1-416-8471752

Received: 13 October 2019; Accepted: 9 November 2019; Published: 12 November 2019

**Abstract:** Claudin-2 is expressed in the tight junctions of leaky epithelia, where it forms cation-selective and water permeable paracellular channels. Its abundance is under fine control by a complex signaling network that affects both its synthesis and turnover in response to various environmental inputs. Claudin-2 expression is dysregulated in many pathologies including cancer, inflammation, and fibrosis. Claudin-2 has a key role in energy-efficient ion and water transport in the proximal tubules of the kidneys and in the gut. Importantly, strong evidence now also supports a role for this protein as a modulator of vital cellular events relevant to diseases. Signaling pathways that are overactivated in diseases can alter claudin-2 expression, and a good correlation exists between disease stage and claudin-2 abundance. Further, loss- and gain-of-function studies showed that primary changes in claudin-2 expression impact vital cellular processes such as proliferation, migration, and cell fate determination. These effects appear to be mediated by alterations in key signaling pathways. The specific mechanisms linking claudin-2 to these changes remain poorly understood, but adapters binding to the intracellular portion of claudin-2 may play a key role. Thus, dysregulation of claudin-2 may contribute to the generation, maintenance, and/or progression of diseases through both permeability-dependent and -independent mechanisms. The aim of this review is to provide an overview of the properties, regulation, and functions of claudin-2, with a special emphasis on its signal-modulating effects and possible role in diseases.

**Keywords:** claudin-2; tight junctions; epithelium; cancer; inflammation; fibrosis; paracellular permeability; proliferation; migration

#### **1. Introduction**

#### *Claudin-2 and the Claudin Family of Tight Junction Proteins*

Epithelial cells maintain tissue homeostasis by forming a protective layer to separate the internal milieu from the outside. The epithelium mediates highly regulated directional transport and thus controls the exchange of water and solutes. Tight junctions (TJ) are found at the apical side of the junctional complexes that connect epithelial cells. Their primary role is to generate cell polarity (referred to as fence function), control paracellular transport (gate function), and provide signaling input for a wide variety of cellular events (reviewed in [1–3]). The TJs are comprised of a multitude of transmembrane and intracellular proteins. The claudin family of TJ membrane proteins consists of 26 (human) or 27 (rodents) small (20-25 kDa) tetraspan proteins expressed in a tissue-specific manner in epithelial and endothelial cells [4–6]. The number of claudins varies between species, and in some species e.g., fish there are a much higher number of claudins, probably due to a larger need for osmoregulation. In this review, we only focus on the mammalian claudin-2. Tsukita and colleagues found that the mouse and human orthologs mostly clustered together, except for claudin-13, which is absent in humans [6–8].

Claudins form the backbone of the TJ strand and determine junctional permeability. Based on permeability properties, members of the claudin family can be classified into channel forming and sealing proteins [2,4,9,10]. The composition of the TJ strand, i.e., the types and ratio of claudins expressed, determines the permeability of the TJs, which can be altered by selectively replacing specific claudins [11].

Claudin-2 was one of the first claudins identified by Tsukita and co-workers as a member of a new class of TJ transmembrane proteins that share tetraspan topology with occludin but have no homology with it [12,13]. Subsequent studies revealed that claudin-2 has a high degree of homology with claudins 1–10, 14, 15, 17 and 19, a group referred to as classic claudins [10]. The mouse and human orthologs of claudin-2 share 70% sequence homology, while their promoters possess a homology of 84% [14,15]. Claudin-2 is a cation-selective channel forming TJ protein expressed in leaky epithelia [16–19]. In this review, we will refer to paracellular channel formation by claudin-2 as its *permeability function*. Importantly, expression of claudin-2 is dynamically modulated by a variety of conditions, and compelling evidence now indicates that altered claudin-2 expression affects vital biological processes, such as proliferation, migration and cell fate decision. These effects cannot be explained by the permeability function of claudin-2, and they appear to be mediated by specific signaling pathways and transcription factors. This newly emerging, incompletely understood role of claudin-2 will be referred to as its signal-modulating function and will be the focus of this review. In fact, such a novel role is in line with the emerging non-classical functions described for several other claudins [20–22].

We will first provide an overview of the properties of claudin-2, and the key inputs regulating its expression. This will be followed by a discussion of its functions including the mounting evidence of its signal-modulating function, and its link to various diseases.

#### **2. Properties of Claudin-2**

#### *2.1. Expression*

Under physiological conditions, claudin-2 mRNA is enriched in the kidneys, where it is exclusively expressed in the proximal tubules, and in the gastrointestinal tract, where the highest expression was reported in the small intestine, liver, gall bladder, and pancreas [23–25]. It is also detectable at lower levels in the stomach and colon [24]. Interestingly, claudin-2 expression was also reported outside of epithelia, e.g., in endothelial cells [26,27] and macrophages (e.g., [28]), but it is not clear whether this represents a physiological localization. As discussed later (see Section 5), pathological conditions affect claudin-2 levels, and may also alter its expression pattern, therefore these observations might indicate pathological expression. Indeed, interleukins and Transforming Growth Factor β1 (TGFβ1) were shown to induce claudin-2 expression in macrophages [28]. Whether claudin-2 has a role in non-epithelial cells remains undetermined.

#### *2.2. Structure and Interactions*

Claudin-2 is a 230 amino acid transmembrane protein, with a calculated molecular mass of 24.5 kDa. Alternatively spliced transcript variants with different 5' untranslated regions have been found for the claudin-2 gene, although their relevance remains unknown. Like other members of the family, claudin-2 is a tetraspan membrane protein consisting of two extracellular domains (ECL1 and ECL2), a small intracellular loop connecting the second and third transmembrane sections and short intracellular N and C terminal portions (Figure 1) ([12,13] and see [5,10] for review on claudin structure). ECL1 (amino acids 29–81), is responsible for paracellular charge selectivity and permeability [29]. The narrow fluid-filled pores of the claudin-2 channel are lined by polar side chains of ECL1 [16,29–32]. Mutagenesis studies revealed that the cation specificity of the channel is due to electrostatic interactions between the transported cations and the negatively charged carboxyl sidechain of D65 and the aromatic residue of Y67 [16,31]. These residues allow cations to permeate in a fully or partially dehydrated state. Intramolecular cis interactions between C54 and C64, two highly conserved cysteines in ECL1 stabilize the TJ pore [33].

**Figure 1.** Claudin-2 structure and interactions with cytosolic multidomain adapters. Claudin-2 consists of two extracellular loops, the longer ECL1 (grey) and shorter ECL2 (black). Claudin-2 also contains 4 transmembrane domains (green box), a cytoplasmic loop (purple), a short N-terminal region and a longer C-terminal region (green). Claudin-2 interacts with the ZO family of TJ plaque proteins (for clarity only ZO1 is depicted) through its PDZ-binding motif - TGYV located at the end of the C-terminus (indicated by the red box). The domains of ZO1 depicted include PDZ1, PDZ2, and SH3, mediating binding of a variety of proteins to create large multiprotein complexes; GUK and A, for actin binding segment. The SH3 domain of ZO-1 was shown to bind to the transcription factor ZONAB, which may play a role in the proliferative effects of claudin-2 (Section 4.2). Afadin is a recently identified claudin-2 partner. The mode of coupling (direct biding or indirect association through adapters) remains to be established. Other newly identified candidate binding partners for claudin-2 include Scrib, Arhgap21, PDLIM2/7, and Rims-2 [34]. The claudin-2 tail contains many potential phosphorylation sites. Among these, Y223 affects the affinity of the PDZ binding domain [35], and S208 appears to be a switch for membrane retention and lysosomal or nuclear localization (see Sections 3.4 and 4.2).

Claudin-2 forms both homo- and heterotypic adhesions, via cis (i.e., molecules in the same cell) and trans (between molecules in neighboring cells) interactions. Cis homodimerization of claudin-2 requires the transmembrane domains and may have a role in organizing the TJ strands [36]. On the other hand, trans homophilic interactions are thought to promote cell-cell adhesion (e.g., [37]). Atomic force microscopy studies probing properties of the interactions between the extracellular loops of claudin-2 revealed that ECL1 was sufficient for homophilic trans interaction, and ECL2 did not mediate these [38]. This is different from what was shown for some other claudins (e.g., claudin-5) [39], where ECL2 had a role in trans interactions, and the reason for such difference remains unknown. Thus, the exact role of the claudin-2 ECL2 is not yet established, since it was not found to be involved in determining permeability or homotypic interactions. Of note, however, a peptide (DFYSP) mimicking a highly conserved region in claudin-2, induced endocytosis, and degradation of the protein [40]. This finding suggests that ECL2 might be key for maintaining claudin-2 at the membrane through a yet to be revealed mechanism.

Some heterotypic trans interactions of claudin-2 have also been described. Claudin-2 can bind to claudin-3 but not to claudin-1 on neighboring cells [41]. However, to date, there is limited information on these interactions, and the exact structural basis for the trans interactions of claudin-2 remains to be mapped. Interestingly, an antagonistic relationship was demonstrated between claudin-2 and two other claudins, and this was implicated in cytokine-induced junction reorganization. Fluorescence recovery after photobleaching (FRAP) analysis of GFP-tagged claudin-4 revealed that claudin-2 and -4 compete with one another for residency at the TJs [42]. Further, claudin-8 was shown to displace claudin-2 from the junctions, resulting in elevated transepithelial resistance (TER) [43].

The intracellular interactions of claudin-2 are pivotal for its signal-modulating functions. However, we only have a limited understanding of the protein network associated with claudin-2. The last four C-terminal amino acids of claudins correspond to a PDZ domain-binding motif, and in classic claudins, the two C-terminal amino acids (YV) are 100% conserved. Specificity for different PDZ domains is determined by variation in the neighboring amino acids close to the PDZ domain-binding motifs [44]. In claudin-2, the last 4 amino acids are T-G-Y-V (N- to C-term). This PDZ-binding motif was shown to associate with a variety of TJ plaque scaffold proteins including the membrane-associated guanylate kinase (MAGUK) family adapters ZO-1-3 [45] (Figure 1). Recently, Tabaries et al. identified several new candidate claudin-2 binding partners [34]. These will be described in more detail in Section 4.4.

The intracellular portions of claudin-2 also undergoes post-translational modifications including phosphorylation [46–48], sumoylation [49], and nitration [50] (see Section 3.4). The effects of posttranslational modifications are not fully understood.

#### **3. Regulation of Claudin-2**

#### *3.1. Context-Dependent Regulation of Claudin-2*

The TJs are continuously remodeled in response to environmental cues. In line with this, claudin-2 expression is dynamically modulated by a multitude of inputs in a context-dependent manner that affect both synthesis and turnover (Figure 2). Protein half-life and trafficking are fine-tuned by a variety of post-translational modifications and interactions with a large array of proteins. Many of the pathways that control claudin-2 are dysregulated in various diseases, resulting in disease-associated changes in claudin-2 expression (see Section 5). In general, studies to decipher the mechanisms controlling claudin-2 were performed in cell lines, and it is important to emphasize that significant context-dependent discrepancies are abundant in the literature. Specifically, some treatments were found to be stimulatory in one cell type, but inhibitory in another. Some of these differences may reflect the complex and likely organ-specific regulation of the protein. In support of this notion, the claudin-2 promoter was affected in the opposite direction by EGF in Caco2 colon cancer and in MDCK tubular epithelial cells [51]. However, some caution is warranted when evaluating data in the literature on claudin-2 regulation in cell lines, as the experiments are impacted not only by difference in origin, culture conditions and passage number of the cells, but also a multitude of other parameters. For example, our studies call attention to the importance of cell confluence and junction maturity as a factor in claudin-2 expression regulation. We found that in LLC-PK1 tubular cells, claudin-2 expression was low in subconfluent cultures and increased as confluence was established [52]. Treatment time is another important parameter that is not always taken into consideration when comparing studies. In tubular and intestinal cell lines the effect of TNFα proved to be opposite depending on the length of treatment [53]. Despite these discrepancies in findings, several mechanisms of claudin-2 regulation are now well established.

**Figure 2.** Schematic overview of claudin-2 regulation and its downstream effects. Various extracellular stimuli activate signaling pathways that impact claudin-2 expression by altering its synthesis, and via post-transcriptional and post-translational modifications. In addition to its permeability effects, altered claudin-2 expression also modulates various cellular processes likely acting via a number of downstream signaling pathways and transcription factors (see Table 1 for details on the signaling and effects downstream from claudin-2).

#### *3.2. Signal Transduction Pathways Regulating Claudin-2 Expression*

A multitude of signaling pathways, many of which are pro-proliferative and cancer-related, were shown to affect claudin-2 abundance. These pathways affect all aspects of claudin-2 regulation, including the control of synthesis via a set of transcription factors, as well as turn-over and localization of the protein. These latter are affected by post-translational modifications and interactions with various proteins (Figure 2). Many pathways affect multiple aspects of claudin-2 control, and the transcriptional effects are often intertwined with the control of turnover. In this section we will summarize the most studied stimuli and pathways that affect claudin-2 expression, highlighting the mode of their action. The transcription factors affecting claudin-2 will be discussed in more detail in Section 3.3.

One of the best-explored regulators of claudin-2 is the epidermal growth factor receptor (EGFR) and its effectors. Both the Ras/Raf/MEK/ERK and the phosphoinositide 3-kinase (PI3K)/Akt/nuclear factor-κB (NF-κB) pathways were found to target claudin-2 [54–56]. EGFR and ERK signaling were shown to affect claudin-2 in many tissues and could mediate effects of a variety of inputs. As mentioned above, both negative and positive effects were reported, highlighting the importance of context (e.g., [51,53,57–62]. EGFR can also mediate effects of other stimuli. For example, histone deacetylases including HDAC4 appear to affect claudin-2 via EGFR and ERK signaling [63,64]. Thus, HDAC inhibitors, that are in use or under consideration for therapy may have a major impact on claudin-2.

The effects of various cytokines on claudin-2 were extensively explored. For example, several interleukins, including IL-1β, 6, 9, 13, and 22 control claudin-2, and some were shown to act via the Jak/Stat pathway [65–67]. The effect of TNFα might be organ-specific, as it was verified to have different effects on claudin-2 in intestinal and tubular cells (e.g., [53,56]).

Kinases, downstream from various receptors may alter claudin-2 synthesis, but direct phosphorylation of the protein could also affect turnover (see Section 3.4). For most of the kinases, these details have yet to be established. Src kinases reduced claudin-2 expression [68,69], while protein kinase C (PKC) family members and the stress kinase JNK augmented it (e.g., [70–73]). Regulation of claudin-2 by Wnt/β-catenin signaling and Rho family small GTPases could also be significant both physiologically and in diseases [52,53,74]. RhoA/Rho kinase and Rac/Pak control both resting claudin-2 levels [52,53,75] and its localization [76].

Finally, several miRNAs, including miR-488, miR-16, and the miR-199a-5p/214-3p gene cluster were also implicated in claudin-2 control [54,77,78]. For example, the miR-199a-5p/214-3p gene cluster was found to be downstream from serum response factor (SRF), a pro-fibrotic transcription factor, and was implicated in high glucose-induced claudin-2 downregulation in peritoneal mesothelial cells undergoing epithelial-mesenchymal transition (EMT) [78].

#### *3.3. Transcription Factors controlling claudin-2 expression*

The TATA-less claudin-2 promoter has binding sites for several transcription factors [14]. Regulation by Caudal-related homeobox (Cdx) homeodomain proteins, hepatocyte nuclear factor (HNF)-1, GATA 2 and -4, AP1, Vitamin D Receptor, TCF/LEF1 and STATs was verified experimentally [14,23,55,72,74,79]. Many of these have significant roles in pathology (see Section 5). HNF-1α and GATA-4 are responsible for claudin-2 expression in specific locations of the gastrointestinal tract, but not in the kidney, highlighting a key organ-specific difference [23,80]. In contrast, GATA-2 was suggested to control claudin-2 in lung and kidney cells [70]. Cdx proteins are intestine-specific transcription factors with significant functions in gut differentiation and carcinogenesis (reviewed in [81]). The claudin-2 promoter is activated by both Cdx1 and 2, which might mediate the effects of the ERK pathway in cancer cells [14,55,82–84]. Interestingly, Cdx factors can collaborate with β-catenin-T cell factor (TCF)/lymphoid enhancer factor (LEF) [74] and HNF-1α [14]. The Cdx1 binding site also contains a functional vitamin D response element, which can directly interact with the Vitamin D Receptor transcription factor [79]. This could be significant for intestinal Ca2+ absorption. Finally, the AP1 complex can mediate the effects of JNK and ERK on the claudin-2 promoter [55,60,72,85], while STATs are effectors of interleukins [65,72].

Many stimuli were shown to decrease claudin-2 levels. This downregulation could be due to inhibition of the above signaling pathways and transcription factors. For example, hyperosmolarity was found to reduce claudin-2 promoter activity by inhibiting PKCβ-dependent GATA-2 transcriptional activity [70]. In addition, transcriptional repressors might also act directly on the claudin-2 promoter. Several EMT-inducing transcription factors are known repressors of tight junction protein genes [86,87]. Accordingly, the transcription factor Snail can induce a drop in claudin-2 [88]. Significantly, the exact mechanisms responsible for claudin-2 downregulation remain mostly unclear. For example, we do not yet have a mechanistic understanding of claudin-2 downregulation in tubular cells (Section 5.3), and the reported negative effects of Src kinases remain unexplained [68,69]. Considering the emerging consequences of reduced claudin-2 expression, a better understanding of these mechanisms will be critical.

#### *3.4. Claudin-2 Turnover, Tra*ffi*cking, and Posttranslational Modifications*

Claudin-2 is a high turnover protein that dynamically cycles between the membrane and intracellular pools [89]. Claudins are connected to both the microtubules and the actomyosin belt through adapter proteins, and these play key roles in assembly and remodeling of the TJs [90–92]. Junctional localization of claudin-2 is not only a prerequisite for permeability functions, but also increases the protein's half-life (e.g., [52,93]). TJ insertion can be promoted or hindered via interactions with membrane and adapter proteins. As mentioned in Section 2.2. claudin-4 and 8 were shown to exclude claudin-2 from the membrane [42,43]. In contrast, ZO family adapters augment the localization of claudin-2 to the TJs. Silencing of ZO-1 and 2 reduced claudin-2 levels, likely in part by enhancing its

degradation [52,67,94,95]. Interestingly, however, loss of ZO-1 also inhibited the claudin-2 promoter, suggesting the existence of feedback regulation between claudin-2 turnover and synthesis [52].

Claudin-2 is retrieved from the TJs by endocytosis. This process was found to be clathrin-dependent in lung and kidney cells [40,48,58]. Other studies have demonstrated a direct binding between claudin-2 and the endocytic scaffold protein caveolin-1 [96,97]. Cytokine-induced gut permeability increase was found to be dependent on caveolin-1 and myosin light chain-dependent endocytosis of TJ proteins [98]. In general, the actomyosin belt plays an essential role in promoting TJ protein retrieval induced by cytokines [91,99], although its specific role in claudin-2 retrieval remains less well established.

The fate of endocytosed claudin-2 is poorly understood. One study found recycling of the endocytosed protein back to the TJs [89]. Several other studies showed that it is targeted for lysosomal degradation (e.g., [40,47,71]). The small GTPases Rab14 and the atypical PKCι, an essential regulator of apico-basal polarity were implicated in the control of claudin-2 trafficking [71]. The autophagy pathway can also target claudin-2 [100,101]. TNFα-induced inhibition of autophagic degradation was suggested to elevate claudin-2 expression in intestinal cells [102]. TNFα also caused an acute ERK-dependent decrease in claudin-2 degradation in tubular cells [53]. Interestingly, similar to some other claudins, claudin-2 was also shown to translocate to the nucleus [103] (see Section 4.2).

Post-translational modifications appear to be key determinants of claudin-2 localization, and turnover, and likely control protein-protein interactions. The claudin-2 intracellular segment contains several potential phosphorylation sites. Phosphorylation of Y224 in human claudin-2 (referred to in some papers as Tyr-6) was shown to affect the affinity of the ZO-1 PDZ binding domain [35], which may impact trafficking. In many publications, the numbering of the position of the C-terminal claudin residues follows the convention that the C-terminal amino acid is referred to as the P0 residue. Subsequent residues toward the N terminus are termed P−1, P−2, etc. Thus, the tyrosine in question is often referred to as position P-6. The possible importance of this site is supported by the fact that many claudins have a tyrosine at this position [35]. S208 is another site that was shown to be phosphorylated. It was implicated in the control of membrane retention, lysosomal localization, and degradation, as well as nuclear localization [46,48,103]. Accordingly, a non-phosphorylatable mutant (S208A) was found to be more cytosolic and lysosomal, while the phospho-mimetic mutant (S208E) localized more to the plasma membrane. Interestingly, mutants, that prevent claudin-2 from insertion to the TJs were found to be poorly phosphorylated, suggesting that TJ insertion is a necessary step for the protein to become phosphorylated. Since the S208 site was implicated both in membrane retention and nuclear localization [46,48,103], it is conceivable that a coupling between claudin-2 TJ insertion and nuclear localization may exist. Accordingly, S208 phosphorylation (e.g., by PKC) could induce plasma membrane localization of claudin-2, while dephosphorylation of this site might promote nuclear translocation [46,48,103].

Claudin-2 also undergoes other post-translational modifications. Sumoylation was found to control membrane localization, ubiquitination, and degradation of claudin-2 [49], and nitration of tyrosine residues was associated with reduced expression [50]. Despite these emerging data, however, we still poorly understand how claudin-2 trafficking and fate are determined.

Importantly, many details of the complex, multilevel regulation of claudin-2 are closely linked to its pathogenic roles, and therefore an improved understanding of the molecular details will be pivotal for further insight into its dysregulation.

#### **4. Functions of Claudin-2**

#### *4.1. Permeability Functions*

Claudin-2 forms paracellular cation and water permeable channels. An elegant study by Weber et al. used a novel trans-TJ patch-clamp technique to show that the claudin-2 channel exhibits symmetrical and reversible conductance of ~90 pS [32]. They found that the channel is gated with one open and two distinct closed states. Claudin-2, on the other hand, is relatively impermeable to

uncharged solutes [17,18], but permeable to cations and water [104]. Interestingly, water transport is mediated by the same pore that allows cation transport ([105] and reviewed in [106]).

These permeability properties are well verified in cell lines and knockout mouse models. Claudin-2 overexpression and silencing in various cell lines decreases and increases TER, respectively [17,18,53, 104,107]. Claudin-2 silencing also eliminates preference of the paracellular pathway towards Na<sup>+</sup> over Cl− [19].

Knockout mouse models also revealed that the selective, site-specific localization and cation channel properties of claudin-2 are essential for highly specialized functions in the kidneys and the bile duct (reviewed in [108]). In the kidney, claudin-2 is pivotal for efficient Na<sup>+</sup> and water reabsorption in the proximal tubules. Accordingly, the S2 segment of the proximal tubules in claudin-2 KO mice showed significantly reduced net transepithelial reabsorption of Na+, Cl−, and water [109]. Although the mice had normal Na<sup>+</sup> homeostasis when fed a regular diet, administration of salt revealed Na<sup>+</sup> handling deficiencies in the kidneys. This suggests that under resting conditions the kidneys could compensate for the loss of the passive claudin-2-mediated Na<sup>+</sup> transport, but under stress, this capacity was insufficient. This conclusion was also supported by findings from Pei et al. [110]. They showed that in claudin-2 knockout mice, upregulation of active transcellular pathways took over the role of passive paracellular Na<sup>+</sup> transport. However, claudin-2 appears to be vital for energy efficiency of the proximal tubular Na<sup>+</sup> reabsorption process, and the higher energy-demand for transcellular transport resulted in medullary hypoxia and increased susceptibility to renal ischemia-reperfusion injury in the claudin-2 KO mice (see also Section 5.3).

In addition to a significant role in ion transport, studies using KO mice also revealed that claudin-2-mediated water transport contributes 23-30% of total water absorption in the proximal tubules [106]. Taken together, the permeability function of claudin-2 in the kidneys is fundamental for Na<sup>+</sup> and water homeostasis and blood pressure regulation [106,109,111].

In the liver and bile system, claudin-2 is highly expressed in hepatocytes and cholangiocytes and plays a central role in water transport associated with bile generation. Accordingly, in claudin-2 KO mice, the rate of bile flow was found to be reduced by half, resulting in a significant increase in bile concentration, and gallstones [112].

In the small intestine, claudin-2 plays a role in luminal Na<sup>+</sup> homeostasis. Surprisingly, however, it was found to be dispensable for Na+-driven nutrient transport [113]. This contrasted with the role of claudin-15, which proved to be indispensable for both luminal Na<sup>+</sup> transport and Na+-driven absorption of vital nutrients. The absence of these two claudins in double knockout mice resulted in severe malnutrition [114].

Taken together, claudin-2 plays a key role as a paracellular permeability molecule. However, as discussed in the next sections, emerging evidence suggests that its functions go beyond permeability.

#### *4.2. Role in Proliferation*

Evidence is mounting that claudin-2 is not only a paracellular channel protein, but also acts as a signal modulator and integrator (Table 1). This conclusion is derived in part from the correlation between claudin-2 expression and altered outcomes, and in part from overexpression and silencing studies that reveal the effects of altered claudin-2 expression on various outcomes. Importantly, dysregulation of claudin-2 expression appears to be an important event in several diseases (see Section 5).

As discussed in Section 3.2, many proliferative pathways affect claudin-2 expression, and these are often dysregulated in diseases, including in cancer. In fact, a strong correlation has been established by many studies between proliferation/cell viability and altered claudin-2 expression. For example, the silencing of the TJ adapter protein cingulin in kidney tubular cells simultaneously elevated claudin-2 levels and caused a RhoA-dependent increase in G1/S phase transition [75]. In A549 lung adenocarcinoma cells, ephrinA1 reduced both claudin-2 expression and proliferation, and both were mediated by Cdx-2 [115]. Recently, miR-488 was shown to suppress both claudin-2 expression and cell viability in colorectal carcinoma cells [77].

These correlations, however, do not prove a causative connection. More definitive proof is provided by gain- and loss-of-function experiments in a variety of cell lines, showing that primary claudin-2 expression changes can alter proliferation (e.g., [67,103,116]). The experimental evidence connecting claudin-2 with downstream effects is summarized in Table 1. Experimental interventions that reduce claudin-2 expression in lung cancer cells reduce proliferation (see Section 5.1) [64,117]. Thus, claudin-2 might be a permissive factor for proliferation and may act as a mediator of pro-proliferative signaling. Accordingly, silencing claudin-2 in the colon cancer cell line Caco-2 and in tubular cells prevented EGF- and TNFα-induced increase in cell proliferation [51,67].

Claudin-2 impacts cell cycle progression and expression of cell cycle regulators both in cultured cells and in a transgenic mouse model [67,103,116]. In a Villin-claudin-2 transgenic mouse model that overexpresses claudin-2 in the colon, the proliferation of colon cells was augmented by claudin-2 overexpression via the PI-3Kinase/Bcl-2 pathway [116]. In lung cancer cells claudin-2 controls the G1/S transition through cyclin D1 and E1 [103]. This effect was mediated by the Y-box transcription factor Zonula occludens 1-associated nucleic acid-binding protein (ZONAB). ZONAB is known to shuttle between the TJs and the nucleus. Interestingly, claudin-2 was also found in the nucleus where it formed a complex with ZO-1 and ZONAB [103]. Moreover, its nuclear levels increased during the G1/S transition. Interaction between claudin-2 and ZONAB was also demonstrated in colon cancer cells, where symplekin, a transcriptional regulator that cooperates with nuclear ZONAB, was shown to control claudin-2 levels [118]. The molecular underpinning of the nuclear translocation of claudin-2 remains insufficiently understood. No classic nuclear localization sequence is recognizable in claudin-2, but mutagenesis studies revealed that the C-terminal cytosolic segment (but not the PDZ domain) was essential for nuclear transport. As discussed in Section 3.4 S208 phosphorylation is a key regulator of nuclear localization, and a phospho-incompetent mutant (S208A) was found to be more nuclear.

Another cell cycle regulator affected by claudin-2 is p27kip1 (cyclin-dependent kinase inhibitor 1B, CDKN1B). The expression of this molecule is high in quiescent cells, where it blocks cell cycle entry. We found that claudin-2 silencing elevated p27kip1 levels, and this effect was mediated by the GEF-H1/RhoA pathway [67]. Since p27kip1 has tumor suppressor effects, its regulation by claudin-2 might be relevant for carcinogenesis.

Altered claudin-2 expression affects several other transcription factors too. Sp1, a zinc finger transcription factor belonging to the Sp/KLF family and c-jun, part of the AP1 early response complex were found to be affected by claudin-2. Specifically, claudin-2 knockdown decreased the levels of phosho-c-Jun and reduced nuclear Sp1 [119,120]. Importantly, these transcription factors also regulate cell proliferation (e.g., [121]) suggesting another possible link.

Recently, claudin-2 was shown to affect miRNAs. This effect was implicated in the control of self-renewal in colon cancer stem cells [122]. Although the mechanisms that connect claudin-2 to miRNA regulation remain undefined, such effects might link claudin-2 to many downstream events.

Finally, yet another intriguing observation is that claudin-2 can localize along the primary cilium [123]. Claudin-2 is dispensable for ciliogenesis, and its exact role in the cilium remains unclear. However, since the primary cilium is present only in quiescent cells and its disassembly is necessary for cell proliferation [124], this intriguing observation clearly warrants further exploration.

#### *4.3. Migration*

The effects of claudin-2 on cell migration may be relevant in both cancer biology and tissue regeneration. Although many studies found that claudin-2 affects migration, significant contradictions are notable among these (Table 1). Specifically, both claudin-2 overexpression and silencing were shown to enhance migration, pointing to context-dependent factors.

In support of a role for claudin-2 in augmenting motility, in various cancer cells claudin-2 overexpression correlates with enhanced migration. In gastric cancer cells, the Helicobacter pylori virulence factor cytotoxin-associated gene (CagA) augmented both claudin-2 expression and invasiveness [82] (see Section 5.1). In line with a stimulatory role, in several cells, a decrease in claudin-2 abundance reduced migration. Non-steroidal anti-inflammatory drugs (NSAIDs), that may provide protection against cancer, reduced both migration and claudin-2 expression in colon, stomach, and lung cancer cell lines [125]. In these studies, the authors provided further evidence for a causal link, by showing that overexpression of claudin-2 stimulated migration and that claudin-2 re-expression restored migration following drug treatment.

Underlying mechanisms for a positive effect of claudin-2 on migration are yet to be unraveled, but a few possible players have emerged. Matrix metalloproteases (MMP) might be important mediators of the effect. This family of zinc-dependent proteases contains over 25 members. Several of them were shown to aid migration through the digestion of extracellular substrates [126] and are considered as potential therapeutic targets. Claudin-2 was shown to affect MMP9 activity, although surprisingly claudin-2 decrease and increase exerted a similar augmenting effect [120,127]. Inhibition of A549 cancer cell migration in a wound-healing assay by claudin-2 knockdown was attributed to reduced MMP-9 expression and activity due to Sp1 inhibition [120]. However, this might not be a universal effect, as claudin-2 overexpression-induced augmented migration of Caco-2 cells proved independent of MMPs [128].

As opposed to the above-described effects, some studies have found that claudin-2 exerts an inhibitory effect on migration. In MDCK tubular cells inducible knockdown of claudin-2 increased MMP-9 mRNA and activity, and this enhanced migration in a wound-healing assay [120]. Similarly, hyperosmotic stress-induced decrease in claudin-2 expression in the same cell line led to augmented migration, and the phenotype was rescued by re-expression of claudin-2 [70]. Of note, MDCK cells are not derived from cancer, but from normal renal tubular epithelial cells, which might explain the different effects on migration. Claudin-2 loss was also shown to augment migration downstream from the transcription factor Spi-B. This factor is normally restricted to the lymphocyte lineage, but it is frequently expressed in lung cancer. Further, it was shown to repress the transcription of claudin-2, which led to enhanced invasiveness [129]. Another study suggested that in osteosarcoma cells claudin-2 loss augmented migration via the afadin/ERK pathway [130]. Along the same lines miR-488 was shown to regulate invasion and lymph node metastasis in colon cancer cells through claudin-2-dependent control of the MAPK pathway [77].

Another mechanism whereby claudin-2 might affect migration is through effects on EMT [131,132]. This process involves a coordinated genetic reprogramming, during which cells lose their epithelial characteristics and gain mesenchymal properties, including enhanced motility. This process is key during embryonic development and wound healing, but also plays a role in cancer metastasis formation. Tight junction downregulation is a hallmark of full-blown EMT [86], but interestingly, multiple studies now suggest a reciprocal correlation too, namely active EMT-regulating roles for claudins. For example, claudin-1 overexpression was shown to promote EMT, while claudin-3 suppressed it [133,134]. While the exact role of claudin-2 in this process remains to be fully established, we recently found that claudin-2 silencing enhanced RhoA-mediated activation of Myocardin-Related Transcription Factor (MRTF), and upregulated Slug, two pro-EMT transcription factors [67]. The potential role of these in altered migration following claudin-2 loss remains unknown.

Taken together, claudin-2 was shown to have both a positive and a negative effect on migration, suggesting a need for optimal claudin-2 levels for efficient migration regulation. Clearly, further mechanistic inquiries will be required to establish the context-dependent mechanisms and importance of these effects.





#### *4.4. Signal-Modulating E*ff*ects of Claudin-2: Emerging Mechanisms Underlying Roles in Biological Processes*

As discussed in Sections 4.2 and 4.3, evidence is mounting that claudin-2 affects cell behavior (Table 1). However, a coherent picture of the underlying mechanisms has yet to emerge. Many studies suggest that claudin-2 may be a signal modulator and integrator protein. Nevertheless, an in-depth understanding of how claudin-2 is linked with altered signaling pathways remains elusive.

The TJ cytoplasmic plaque contains a large array of signaling and regulatory proteins. These are connected to the TJ membrane proteins by various adapters. For many of these proteins, recruitment to the junctions results in inactivation, or a spatial restriction of activity (for review see [90,136,137]). The specific role of individual claudins in the organizing of the cytoplasmic plaque is not yet resolved. As mentioned in Section 2.2, the cytoplasmic tail of claudin-2 interacts with multiple proteins, but the claudin-2 interactome has not yet been fully mapped, and context-dependent regulation of the interactions remains unknown. As mentioned earlier, a recent study from the Siegel group identified some candidate binding partners of the claudin-2 PDZ domain [34]. Using metastatic breast cancer cells with elevated claudin-2 expression, they demonstrated that the last three C-terminal amino acids of claudin-2 (comprising the PDZ-binding motif) are necessary for anchorage-independent growth. They identified several potential partners of this motif, including afadin, which is an actin filamentand Rap1 small GTPase-binding protein encoded by the MLLT4/AF-6 gene. Loss of afadin impaired colony formation of breast cancer cells in soft agar and reduced lung and liver metastasis, verifying the significance of the claudin-2/afadin complex. Importantly, some of their findings suggested the existence of afadin-independent mechanisms too. The identified potential other partners, including the polarity protein Scrib, the small GTPase regulator Arhgap21, PDZ-LIM domain-containing proteins (PDLIM) 2 and 7, and the exocytosis-controlling Rims-2 (regulating synaptic membrane exocytosis protein 2) may mediate such downstream effects. However, whether these are direct binding partners or interact via adapters, and their potential functional significance remains to be established.

The crosstalk between claudin-2 and the cytoskeleton may play a central role in mediating downstream effects. Claudin-2, as all TJ membrane proteins, is anchored to the junctional actomyosin belt and microtubules through adaptor proteins. Among these, ZO family proteins directly bind claudins, and connect to F-actin either directly, or indirectly through actin-binding proteins [138,139]. Thus, it is conceivable that altered claudin-2 expression might affect the organization of the prejunctional cytoskeleton. In support of this, we recently found that in tubular cells, claudin-2 silencing altered acto-myosin organization through RhoA [67].

The microtubules also play a role in junction regulation and might be affected by the TJ organization. Cingulin and its paralog paracingulin interact with TJ proteins, likely through ZO adapters, and thus connect to actin. Cingulin also binds the microtubules and has a role in organizing the microtubular planar apical network in mammary epithelial cells [140,141]. Further, association of cingulin with actin filaments and the microtubules was found to be regulated by phosphorylation and plays a crucial role in the control of the epithelial barrier [141]. The specific roles of individual claudins as organizers of the cytoplasmic plaque complexes and their impact on the cytoskeleton are yet to be uncovered. Nevertheless, it is conceivable that some of the signal modulating effects of claudin-2 is mediated by an impact on the acto-myosin or the MTs.

Interestingly, claudin-2 appears to have a bidirectional relationship with the cytoskeleton and some signaling pathways. Specifically, some of the pathways modulated by claudin-2 can also regulate its expression. This suggests the existence of various regulatory feedback cycles (Figure 2). Such regulatory loops also pose a challenge for defining clear "upstream" and "downstream" events. The ERK and Rho signaling pathways are prominent examples of this complexity. The role of MEK/ERK signaling in basal and stimulus-induced claudin-2 expression is well documented [53,55,60,61,85,142,143]. Besides, a recent study reported the converse effect: claudin-2 silencing was found to augment MEK/ERK1/2 signaling [130], an effect that may help to restore claudin-2 levels. A complex relationship also exists between Rho family small GTPases and claudin-2. We found that RhoA/Rho-kinase are negative regulators of claudin-2 expression in cultured tubular cells [53]. On the other hand, as mentioned above, in the same cells claudin-2 silencing enhanced RhoA activity through the exchange factor GEF-H1, suggesting that in resting cells, claudin-2 might suppress RhoA activation [67]. Thus, claudin-2 loss can augment Rho activity, which in turn may further reduce claudin-2 expression, in a possible self-augmenting cycle.

#### **5. Claudin-2 in Diseases**

A growing number of studies document altered claudin-2 expression, phosphorylation and/or localization in various pathological conditions. Initially, the described claudin expression alterations were considered an epiphenomenon, i.e., they were assumed to arise due to the underlying disease process but were thought not to be causally contributing to pathogenesis. Nevertheless, interest in pathological alterations in claudin-2 expression was boosted by the hope that it can be used as a diagnostic and/or prediction marker. As described in Section 4, in the past years, strong evidence accumulated in support of a causal link between claudin-2 dysregulation and functional alterations, the key points of which are the following. First, signaling pathways that are known to be overactivated in diseases can alter claudin-2 expression, and a good correlation exists between disease stage and claudin-2 expression. Second, loss-of-function and gain-of-function studies recapitulate some aspects of the functional changes. Thus, primary changes in claudin-2 expression can alter cell behavior. The mounting experimental evidence prompted a paradigm shift, favoring a pathogenic role for claudin-2. Although dysregulation of claudin-2 is likely not a primary cause, pathological changes in claudin-2 abundance and/or localization might play significant roles in the generation, maintenance and/or progression of diseases.

In the following sections, we will provide an overview of the evidence linking claudin-2 to cancer, and various non-malignant pathologies, such as inflammatory gastrointestinal and kidney diseases.

#### *5.1. Claudin-2 in Cancer and Metastasis Formation*

An expanding body of literature documents dysregulated claudin-2 expression in gastric, colorectal, lung, breast, and renal carcinomas and in osteosarcoma (e.g., [51,144–148]. As described in Sections 4.2 and 4.3, claudin-2 levels affect processes underlying carcinogenesis and metastasis formation, including proliferation, migration and epithelial-mesenchymal transition. The following overview however also highlights that the role of claudin-2 is likely complex, and differences in its impact might exist not only based on tissue origin but also cancer stage.

*Cancers of the gastrointestinal tract*—Claudin-2 is highly expressed in gastric and colorectal cancers and its expression level shows a good correlation with the development of these tumors. For example, a gradual increase was observed in claudin-2-positivity in various stages of gastric carcinogenesis from no expression in gastritis to elevated expression in dysplasia and gastric intestinal-type adenocarcinoma [149]. In fact, in this study, 73% of adenocarcinoma samples were found to be claudin-2 positive.

As described in Sections 3.2 and 3.3, claudin-2 is the target of key signaling pathways and transcription factors central to gastrointestinal cancers. One prominent example is the significant correlation between Cdx2 and claudin-2 expression in gastric dysplasia and cancer [150]. Further, a potentially important link between claudin-2 and H. pylori was also uncovered. H. pylori infection shows a strong correlation with gastric cancer [151], and the virulence factor CagA is considered a gastric oncogene. During infection, this protein is translocated into and reprograms the gastric cells. Interestingly, CagA augmented claudin-2 expression in gastric cells, and promoted invasiveness through effects on Cdx2 [82].

Hyperactivation of Wnt signaling and the consequent increase in gene transcription by TCF/ LEF is a hallmark of colon cancer (reviewed in [152]). Wnt-1 was shown to increase claudin-2 promoter activity through β-catenin/LEF-1 [74]. Increased claudin-2 and β-catenin expression were detected in active inflammatory bowel disease (IBD), adenomas, and IBD-associated dysplasia, but not in acute, self-limited colitis [153]. These data suggest that β-catenin might mediate an increase in claudin-2 during carcinogenesis.

EGFR signaling is yet another key pathway linking colon cancer and claudin-2. EGFR activation upregulated claudin-2 in colon cancer cells, and claudin-2 expression was decreased in the colon of waved-2 mice that have EGFR activation deficiency [51] (see also Section 3.2 on the effects of EGFR on claudin-2).

Taken together, these studies suggest that the initial upregulation of claudin-2 could be due to the overactivation of the above-discussed pathways, and this may promote carcinogenesis. To prove a causal link between elevated claudin-2 and the properties of cancer, exogenous claudin-2 was overexpressed in colorectal cancer cell lines. An increase in claudin-2 expression promoted colonocyte proliferation and anchorage-independent colony formation and stimulated tumor formation in colorectal cancer xenografts (e.g., [51,116]). Further, claudin-2 expression was a negative predictor for post-chemotherapy disease-free survival of colon cancer patients [51].

Another key role of claudin-2 could be its effect on colorectal cancer stem-like cells [122]. Claudin-2 promoted self-renewal of these cells via miRNAs and the hippo effector yes-associated protein (YAP). Interestingly, claudin-2 was also detected in human colorectal cancer-associated fibroblasts (CAFs). This finding is especially intriguing since claudin-2 is regarded as an epithelial molecule. Its expression in CAFs was linked to KRAS mutation status and correlated with reduced progression-free survival [154]. Of note, claudin-2 was also shown to be enriched in macrophages associated with mammary tumors [28]. Thus, an attractive hypothesis is that claudin-2 may increase proliferation/survival and migration of fibroblasts and macrophages in the tumor environment.

*Lung cancer*—Primary lung cancer is among the deadliest types of tumors worldwide, and adenocarcinomas are the most frequent forms of non-small cell lung cancer. Claudin-2 expression in normal bronchial epithelium is low [85]. In contrast, according to one study, two-thirds of examined lung adenocarcinoma samples overexpressed claudin-2 [145]. Surprisingly, in these cells, claudin-2 was found mostly in cytoplasmic granules. Adenocarcinoma is often associated with upregulated EGFR activity and KRAS/MEK/ERK signaling. Indeed, EGFR signaling is key in upregulating claudin-2 in the human lung adenocarcinoma cell line A549 [60]. These cells were used in a series of studies to demonstrate that claudin-2 may be a therapeutic target in lung cancer. The studies showed that rapid proliferation of A549 cells required claudin-2. Accordingly, claudin-2 knockdown impaired cell growth and migration [103,120] and elevated sensitivity for anti-cancer agents [119]. Importantly, a variety of interventions, including epigenetic inhibitors and various flavonoids reduced claudin-2 expression and decreased proliferation. These effects were counteracted by retransfection of claudin-2 [64,117]. Claudin-2 abundance can also be reduced by a peptide that mimics a sequence in ECL2. This peptide-induced endocytosis and lysosomal accumulation of claudin-2, resulting in lysosomal damage and necrotic cell death [40]. Taken together, these studies provided strong evidence that targeting claudin-2 can mitigate proliferation and raised hope that this approach will prove useful in cancer therapy (see further elaboration in Section 5.4).

*Osteosarcoma and breast cancer*—Interestingly, in some tumors, claudin-2 expression is decreased. Osteosarcoma cells have lower claudin-2 expression than normal osteoblasts [130], and this is associated with enhanced migration. In breast cancer, increased or decreased claudin-2 expression might be associated with different cancer types and stages [155,156]. As discussed below, in this type of tumor, claudin-2 expression appears to correlate with metastasis formation.

*Metastasis*—The importance of claudin-2 in metastasis is also emerging. Claudin-2 might affect migration via a variety of effects. Its possible role in migration was discussed in Section 4.3. Beyond general effects on migration, claudin-2 also appears to affect metastasis through mediating specific adhesion of the metastatic cells, thereby promoting invasion into distant organs. Interestingly, the exact role of claudin-2 in metastasis appears to vary based on the site of invasion. In some breast cancer reduction in claudin-2 was associated with lymph node metastasis and higher clinical stage [156]. In contrast, claudin-2 overexpression was shown to augment breast carcinoma metastasis to the liver. Claudin-2 expression is elevated in liver metastases compared to primary breast cancer cells [157]. Indeed, claudin-2 promoted attachment and survival of metastatic cells specifically in the liver and enhanced their interactions with hepatocytes [37]. Due to these effects, claudin-2 was found to be a negative prognostic factor that predicts liver metastasis in breast cancer [158]. Metastatic breast cancer cells showed a general increase in adhesion to the extracellular matrix, as claudin-2 promoted surface expression of α2β1- and α5β1-integrins [157]. However, while these effects might contribute to invasiveness in general, enhanced interaction between hepatocytes and breast cancer cells proved to be independent of integrins. Instead, the effect required the first claudin-2 extracellular loop (ECL1) and was likely attributable to homotypic trans interactions between claudin-2 molecules. Thus, claudin-2 in the metastatic breast cancer cells can bind to claudin-2 on hepatocytes, thereby promoting their adhesion to the liver [157]. Further, the PDZ-binding motif (YV) in claudin-2 was found to be necessary for anchorage-dependent growth and metastases [34]. These exciting findings raise hope that blocking peptides or neutralizing antibodies against the identified domain might be effective in reducing liver metastasis (see Section 5.4).

#### *5.2. Claudin-2 in Gut Inflammation*

Claudin-2 emerged as an important factor in IBD, a term describing conditions such as ulcerative colitis and Crohn's disease. These chronic diseases are hallmarked by compromised epithelial barrier and are thought to be caused by a dysregulated immune response in the gut (e.g., reviewed in [159,160]). Increased paracellular permeability is likely a critical pathogenic factor. Specifically, the disease appears to be maintained by a vicious cycle: increased epithelial permeability due to dysregulation of TJ proteins elicits an aberrant immune response, and the ensuing inflammation further disrupts TJs and aggravates the condition. The effects of inflammation on epithelial permeability have been long known. Elevated claudin-2 expression in IBD is attributed to the presence of cytokines (e.g., [56,161,162] and reviewed in [159,163]). The permeability function of claudin-2 appears to be central in IBD, although a causal link is hard to decipher. Studies in transgenic mice explored and verified the role of the permeability function of claudin-2. In contrast, little is known about the possible significance of claudin-2 as a signal modulator in IBD. Somewhat surprisingly, recent studies point to a possible protective role of claudin-2 against injury. A transgenic mouse with targeted overexpression of claudin-2 in the colon was protected against colitis-associated injury [116]. This study also highlighted the complex roles of claudin-2 in intestinal homeostasis and IBD. Claudin-2 overexpression in the colon augmented mucosal permeability but did not by itself cause inflammation. Instead, the mice had longer colons and elongated crypts, a result of accelerated colonocyte proliferation. Most notably, despite the leaky colon, the mice were significantly protected against experimental colitis, and inflammation-associated gene expression was sharply downregulated. Some of the effects could be the result of reduced apoptosis and augmented regeneration due to faster proliferation. Thus, in this respect, claudin-2 elevation may at least initially confer some protection in IBD. This notion is also supported by another study that used claudin-2 null mice to explore the effects of intraperitoneally injected TNFα and experimental colitis [135]. The data revealed augmented colorectal inflammation in the KO mice compared to WT animals. These effects were mediated by NF-κB signaling. The mice also had increased expression of IL-6 and IL-1β and higher intestinal myosin light chain kinase levels. Overall, these studies suggest that increased claudin-2 expression might suppress inflammatory signals. Importantly, although the proliferative effect observed could be beneficial for regeneration, this is a double-edged sword. IBD elevates the risk for developing colorectal cancer and augmented claudin-2 expression could contribute to this. As mentioned in Section 5.1, claudin-2 levels rise with the development of dysplasia and cancer. This raises the intriguing possibility that elevated claudin-2 expression might be a crucial connection between inflammation and cancer. Thus, while elevated claudin-2 might protect against injury and enhance regeneration, it could raise the risk of cancer.

The role of claudin-2 in intestinal diseases might go beyond IBD, as it was also implicated in diseases caused by enteropathogenic bacteria. Salmonella invasion elevated claudin-2 protein and mRNA levels both in cultured cells and in the intestine of mice. Upregulation of claudin-2, in turn, elevated permeability and promoted internalization of the bacteria [73]. Finally, recent studies raise the

possibility that claudin-2 might contribute to chronic pancreatitis, a progressive inflammatory disease that is frequently associated with alcohol consumption. A genome-wide association search found that polymorphisms of the claudin-2 locus confer an increased risk of alcohol-induced pancreatitis [164]. Since claudin-2 is essential for bile formation, it will be interesting to see if claudin-2 polymorphism might also predispose for gall bladder disease and gallstones.

#### *5.3. Claudin-2 in Kidney Disease*

Claudin-2 mRNA levels are highest in the kidney, where it is a key mediator of cation and water transport in the proximal tubules (reviewed in [165–167]). Although a direct role of claudin-2 in kidney disease has not been definitively established, several observations suggest a possible link.

Claudin-2 is affected by inflammatory cytokines in tubular cells. Specifically, short-term TNFα treatment caused an increase in claudin-2 abundance [53]. In contrast, prolonged (>8h) TNF-α or IL-1β treatment downregulated claudin-2 expression [53,67]. Interestingly, a large variety of potentially harmful stimuli can downregulate claudin-2 expression in tubular cells. In addition to these cytokines, metabolic acidosis [168], changes in osmolarity [48,127], oxidants [169], EGF [58,170], sheer stress [171], the flavonoid quercetin [172], and drugs, such as the immunosuppressants sirolimus and cyclosporine A [173], reduce claudin-2 levels. Reduced claudin-2 expression was also recorded in various kidney disease animal models, including cisplatin-induced nephrotoxicity [174], diabetic nephropathy [50] and obstructive nephropathy-induced fibrosis [67]. These findings are of significance because claudin-2 was found to exert a protective effect against kidney injury by reducing the energy demand of transport processes [110]. Further, as mentioned above loss of claudin-2 in tubular cells induced RhoA activation, and claudin-2 knockout mice had elevated RhoA activity [67]. This led to reduced proliferation and activation of MRTF, a transcription factor that mediates profibrotic epithelial transition [67,175]. Thus, reduced claudin-2 expression in stress conditions might increase susceptibility for injury and may promote fibrosis.

Finally, the proximal tubular paracellular pathway also plays a central role in Ca2+ reabsorption. Accordingly, claudin-2 null mice were hypercalciuric, a condition that enhances kidney stone formation [109]. Thus, conditions that reduce the abundance of claudin-2 in the kidney might also lead to kidney stone disease, a condition affecting a significant portion of the population.

#### *5.4. Development of Therapies Targeting Claudin-2*

Several studies raise hope that targeting claudin-2 could have beneficial effects. Indeed, in a proof of principle study, a monoclonal antibody that recognizes the first extracellular loop (ECL1) of claudin-2 was shown to prevent TNFα-induced TJ disruption [176]. The same group recently generated a human-rat chimeric monoclonal antibody against claudin-2 and showed that it accumulated in claudin-2 positive sarcoma xenografts. The antibody also suppressed tumor growth [177]. Importantly, no major adverse effects were found.

Taken together, accumulating evidence supports that claudin-2 can be considered a diagnostic and prognostic marker in various diseases and is an exciting potential therapeutic target.

#### **6. Open Questions**

Claudin-2 is now established as a pathogenic factor in several diseases due to its permeabilitydependent and -independent functions. The recognition that claudins have major signal modulator, permeability-independent functions reenergized claudin research, leading to an explosion in the number of studies published. Although the fact that altered claudin-2 expression affects signaling and cell behavior is firmly established, many questions remain unanswered. It is worthwhile to articulate some of these as they set the direction for future research. Importantly, our understanding of mechanisms mediating pathological dysregulation of claudin-2 is still incomplete, which precludes the design of interventions. There are also fundamental unanswered questions regarding the mechanisms that link claudin-2 to various downstream events. We do not know the role of claudin-2 localization in its signal-modulating effects. Are these effects mediated by TJ localized claudin-2? Given that claudin-2

was shown to reside in other compartments (cytosolic vesicles, nucleus), some of the non-classical effects may well be mediated by these claudin-2 sub-pools. Further, many studies revealed the importance of altered claudin-2 expression in the downstream effects, but we do not know whether these are exclusively related to the number of molecules or if further regulation (e.g., by distinct posttranslational modifications) are also involved. Extra- and intracellular binding partners of claudin-2 are likely crucial mediators of signal-modulating effects. Indeed, TJ-localized claudin-2 can interact with other membrane proteins and may act as a sensor of neighboring cells. However, it is not clear whether claudin-2 needs to be engaged for its signal-modulating effects. Further, the claudin-2 interactome is only now starting to emerge and uncovering the context-dependent regulation of interactions will be key for a better understanding. Adapters in the TJ plaque generate signaling complexes, that may change depending on the availability of claudin-2, leading to recruitment or release of signaling intermediates and alterations in the cytoskeleton organization. However, the specific role of claudin-2 in such events remains unknown. Ongoing research from many groups will likely help fill these knowledge gaps leading to a better mechanistic understanding of the signal-modulating effects of claudin-2. The emerging knowledge has already informed the design of interventions targeting claudin-2. Such future therapies have the potential to benefit many patients with a broad spectrum of diseases.

**Author Contributions:** Writing–Original Draft Preparation, S.V., S.A., K.S.; Review & Editing, S.V., S.A., K.S.; Visualization, S.A.; Supervision, K.S.; Funding Acquisition, K.S.

**Funding:** This research was funded by the KIDNEY FOUNDATION OF CANADA, the CANADIAN INSTITUTES OF HEALTH RESEARCH (CIHR), grant number: PJT-149058 and MOP-142409; and NATURAL SCIENCES AND ENGINEERING RESEARCH COUNCIL OF CANADA (NSERC) grant number: RGPIN-2017-06517. Shaista Anwer received scholarships from the Research Training Center of the St Michael's Hospital and University of Toronto through an open fellowship.

**Acknowledgments:** The authors would like to express their gratitude to Andras Kapus and Mirjana Jerkic for critical reading of the manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Abbreviations**


#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
