Emerging Multifunctional Roles of Claudin Tight Junction Proteins in Bone

Abstract

The imbalance between bone formation and resorption during bone remodeling has been documented to be a major factor in the pathogenesis of osteoporosis. Recent evidence suggests a significant role for the tight junction proteins, Claudins (Cldns), in the regulation of bone remodeling processes. In terms of function, whereas Cldns act “canonically” as key determinants of paracellular permeability, there is considerable recent evidence to suggest that Cldns also participate in cell signaling, ie, a “noncanonical function”. To this end, Cldns have been shown to regulate cell proliferation, differentiation, and gene expression in a variety of cell types. The present review will discuss Cldns' structure, their expression profile, regulation of expression, and their canonical and non- canonical functions in general with special emphasis on bone cells. In order to shed light on the noncanonical functions of Cldns in bone, we will highlight the role of Cldn-18 in regulating bone resorption and osteoclast differentiation. Collectively, we hope to provide a framework for guiding future research on understanding how Cldns modulate osteoblast and osteoclast function and overall bone homeostasis. Such studies should provide valuable insights into the pathogenesis of osteoporosis, and may highlight Cldns as novel targets for the diagnosis and therapeutic management of osteoporosis.

Cell-to-cell interaction is important in the development and maintenance of various biological tissues. The cell junctional complex consists of four types of proteins including gap junctions (eg, connexins), hemidesmosomes (eg, integrins), adherens (eg, cadherins), and tight junctions (1). In the skeleton, it has been shown that bone cells form a variety of intracellular junctions (2, 3), and there is substantial evidence demonstrating the importance of gap junctions in skeletal development; for example, connexin 43-deficient mice exhibited reduced bone mineral density (BMD) and skeletal abnormalities (2). Regarding tight junctions, several investigators have documented the ability of bone cells (osteoblasts and osteocytes) to form tight junctional structures (4–6). Tight junction structures, junctional strands, are composed of several types of transmembrane proteins including occludin, junctional adhesion molecule, tricellulin, and Claudins (Cldns) (1). It was previously reported that occludin is expressed in osteoblasts and its targeted disruption in mice resulted in cortical bone thinning and postnatal growth retardation (7, 8). Of the various tight junction proteins, extensive evidence suggests that Cldns are the primary proteins responsible for the formation of tight junctional strands, and there is also evidence showing that they participate in intracellular signaling that controls cell proliferation and differentiation in a variety of cell types (1, 9–11). In this regard, recent studies suggest a significant role for Cldns in regulating bone remodeling via their noncanonical actions. Thus in this review, we discuss the Cldns' structure, regulation, functions, and their mechanisms of action in general with special emphasis in bone biology. The notion that Cldns exert multifunctional roles in various tissues including bone should uncover new avenues of research that lead to the elucidation of potential roles for Cldns in the diagnosis and treatment of skeletal diseases.

Structure

The Cldn family of protein molecules consists of 27 members that have been identified in mouse and human cells, with a molecular weight ranging from 20–34 kDa (12). In terms of structure, Cldns are composed of four transmembrane domains, two extracellular loops, a short intracellular loop, and cytosolic amino and carboxy termini (Figure 1) (13). Whereas the amino acid sequence of the first and fourth transmembrane domains is highly conserved among the different Cldns, that of the second and third domains is not (14). Structure-function studies revealed that Cldns' first extracellular loop contains several charged amino acids that have been found to play an important role in determining their paracellular barrier charge selectivity (1, 15). In addition, the first extracellular loop has two signature sequences including a G-L-W motif and two highly conserved cysteines, which are thought to confer an important function in stabilizing the protein structure (9, 12, 16). The second extracellular loop is folded in a helix-turn-helix motif that is involved in Cldn-Cldn interactions, and has been found, in some Cldns, to serve as a receptor for the colistridium perfringens enterotoxin (17–20). Regarding the carboxy terminus, it has been shown to possess the highest structural diversity among the Cldns. It also has a PDZ-binding motif that enables Cldns to interact with PDZ domain containing proteins such as zonula occludens (ZO)-1/2/3, MUPP1, PATJ, among others (12, 21). The interactions between Cldns and cytoplasmic scaffolding proteins such as ZO are predicted to be essential for linking the Cldn family of proteins to the cytoskeleton and/or for their participation in intracellular signaling (9, 22). Additionally, the carboxy terminus has phosphorylation sites for Ser/Thr, and Tyr Kinases that serve regulatory roles by modulating the Cldn's localization and function (12). Recently, several phylogenetic trees and classifications are suggested based on the structure in various species, and is comprehensively described by Gunzel et al (23). Based on sequence analysis and functional properties of the mouse variants, Cldns are classified into two major groups: 1, the “classic” Cldns, that include Cldn 1–10, 14, 15, 17, and 19, based on their close similarities; and 2, the “nonclassic” Cldns, which include the remaining Cldns including Cldn-12, 13, 16, 18, 20, 22, and 23, as they seem to be less similar/related (21). However, the issue of whether classic and nonclassic Cldns exert different roles or are regulated differently remains to be studied.

Figure 1.

Schematic representation of Cldns structure. Cldns are composed of four transmembrane domains (I, II, III, and IV), two extracellular loops (ECL1 and ECL2), a short intracellular loop, and cytosolic amino and carboxy termini.

Expression and distribution profile

Cldns exhibit diverse tissue/cell type-specific patterns of expression (Table 1). Some tissues/cells such as the epidermis express several Cldns (24), whereas others such as sertoli cells express one or two Cldns (25). Interestingly, Cldn expression also varies within the same tissue. For example, in mouse kidney, at least 15 Cldns are expressed with distinct expression patterns in every segment of the nephron. Thus, whereas the proximal tubules express Cldn-1, 2, 10, 11, 12, and 14, the distal tubules express Cldn-3, 7, 8, 10, and 11 (1). There is increasing evidence that the expression of Cldns is also developmental stage specific (9). In mouse jejunum, the expression of several Cldns is increased or decreased during neonatal development (26). For example, Cldn-19 expression was detected at birth, peaked at day 14, and then was undetectable by day 28. The expression of several Cldns during embryonic development has also been investigated. Thus, during mouse development (between E.7 and E.17), the expression of Cldn-21 and 24 was found to increase progressively, whereas that of Cldn-26 and 27 decreased (9, 27). There is also increasing evidence that the expression of Cldns changes in response to pathological conditions such as cancer and inflammation. Several studies have reported upregulation or downregulation of Cldns in association with various cancers (28). Thus, the known complexity in the expression of patterns of Cldns in different tissues is consistent with both common and Cldn-specific functions among various Cldns.

Table 1.

Cldns Tissue Distribution and Functions

Cldn	Tissue Expression	Canonical Function	Noncanonical Function
1	Ubiquitous: Most epithelial tissues and vascular endothelial (9)	Barrier forming-cation selective (90)	Cell proliferation (78, 91) Cell differentiation (92) Cell motility and migration (during development and in cancer cells) (93–95) Cell signaling and regulating gene expression (96, 97) Anti-apoptosis (98, 99) Hepatitis C virus and dengue virus entry cofactors (100, 101)
2	Typical for epithelial tissues (9, 21)	Pore forming-cation selective (102, 103)	Cell proliferation (104, 105) Cell signaling (106)
3	Epithelial tissues and vascular endothelium (9)	Barrier forming-cation selective (107)	Cell proliferation (108) Cell motility and migration (cancer cells) (109) Gene expression (110) Apoptosis (111) Angiogenesis (111) Receptor for Clostridium perfringens enterotoxin (112)
4	Epithelial tissues (9)	Predominantly barrier forming-cation selective (113)	Cell motility and migration (normal and cancer cells) (109, 115)
		Also act as Cl-pore (114)	Gene expression (110)
			Cell signaling (116)
			Angiogenesis (117)
			Receptor for Clostridium perfringens enterotoxin (112)
5	Typical for vascular endothelium, such as blood brain barrier (39)	Barrier forming-cation selective (119, 120)	Cell motility and migration (cancer cells) (121, 122)
	Some epithelial tissues (118)
6	Embryonic epithelium (123)	Barrier forming-cation selective (126)	Cell proliferation (127, 128)
	Kidney (neonates) (124)		Cell differentiation (129, 130)
	Liver (125)		Cell motility and migration (cancer cells) (127, 128)
			Apoptosis (131)
			Coreceptor for hepatitis C virus (100, 125)
7	Epithelial tissues (9)	Barrier and pore forming (132–134)	Cell proliferation (127, 135) Cell motility and migration (cancer cells) (127, 135) Gene expression (136) Apoptosis (135, 137)
8	Epithelial tissues (9)	Predominantly barrier forming-cation selective (138, 139)	Receptor for Clostridium perfringens enterotoxin (140)
9	Kidney (neonates) (124)	Barrier forming-cation selective (126)	Cell proliferation (127)
	Inner ear (141)		Cell motility (cancer cells) (127)
	Liver (125)		Coreceptor for hepatitis C virus (100)
10a	Epithelial tissues, mainly kidney (9)	Pore forming-anion selective (142)	Cell motility and migration (cancer cells) (143)
10b	Epithelial tissues, mainly kidney (9)	Pore forming-cation selective (142)	Cell motility and migration (cancer cells) (143)
11	Sertoli cells (144)	Predominantly Barrier forming-cation selective (146)	Cell proliferation (147)
	Oligodendrocyte and myelin sheath (144)		Cell differentiation (148)
	Inner ear (141)		Cell motility and migration (cancer cells) (149)
	Choroid plexus epithelium (145)		Apoptosis (148)
12	Stomach, intestine, salivary gland, epidermis, urinary bladder, and vascular endothelium (9, 150)	Not well characterized: pore forming-cation selective (33)	Not known
13	Hematopoietic tissues (151)	Not known	Cell proliferation and differentiation (151)
	Colon and urinary bladder (150)
	Kidney (neonates) (124)
14	Kidney and liver (49)	Barrier forming-cation selective (44)	Cell proliferation (152)
	Inner ear (141)		Angiogenesis (152) Receptor for Clostridium perfringens enterotoxin (140)
15	Intestine (150)	Predominantly pore forming-cation selective (146)	Cell proliferation (155)
	Respiratory tract (153)	Also acts as Cl-barrier (15)
	Mammary epithelium (154)
16	Kidney (52) Mammary epithelium (154)	Pore forming-cation selective (156, 157)	Cell motility and migration (cancer cells) (158)
17	Kidney (159)	Pore forming-anion selective (159)	Not Known
18-1	Predominantly in lung (160)	Not known	Not Known
	Kidney (160)
	Inner ear (141)
18-2	Predominantly in stomach (160) Osteoclasts (32) Osteoblasts (36) Inner ear (141) Esophagus (161)	Barrier forming-cation selective (161)	Cell differentiation (32)
19	Kidney (51, 162)	Predominantly barrier forming-anion selective (164)	Not known
	Retinal pigment epithelium (51)	Also act as Na barrier (165)
	Myelin sheath (163)
20	Intestine (155)	Not known	Not known
	Brain capillary endothelial cells (166)
21	Intestine, stomach, liver, and kidney (9)	Not known	Not known
22	Brain capillary endothelial cells (166)	Not known	Not known
23	Intestine (155)	Not known	Not known
	Brain capillary endothelial cells (166) skin, placenta, stomach, and germinal center B-cells (9)
24	Intestine, stomach kidney, and heart (9)	Not known	Not known
25	Intestine, stomach, liver, kidney, heart, and brain (9)	Not known	Not known
26	Intestine and brain (9)	Not known	Not known
27	Intestine and liver (9)	Not known	Not known

Cl indicates chloride.

As for Cldn distribution in bone, in older studies several investigators have documented the ability of bone cells (osteoblasts and osteocytes) to form tight junctional structures (4–6). However, the evidence of the presence of tight junction associated proteins has been recently confirmed (8). Specifically, several Cldns including Cldn-1 to 12, 14 to 20, 22, and 23, were reported to be expressed at the mRNA level in rat osteoblasts (6, 8). In fact, immunohistochemical analysis of decalcified tibial sections revealed that the expression of selected Cldns (Cldn-5, 11, 14, 15, and 16) was localized at the bone lining cells (inactive osteoblasts) (8, 29). Furthermore, a cell border localization of Cldn-1 was shown in an osteoblast-like cell line (30). In another study, Arana-Chavez et al (31) observed the presence of tight junctional structures between osteoblasts in early osteogenesis by freeze fracture and ultrathin section electron microscopy (5). Moreover, Cldn-1, 2, and 6 mRNA expression was found to be higher in osteoblast like MC3T3-E1 cells compared to osteocyte-like MLO-Y4 cells (30). By contrast, the expression of Cldn-1 and 2 in rat osteoblasts was found to be higher in the mineralization stage compared to the proliferation stage, thus suggesting that the expression levels of Cldn-1 and Cldn-2 in rat osteoblasts may be differentiation stage dependent (6). The observed changes in the expression profile of Cldns are consistent with the idea that Cldns play an important role during skeletal development.

In terms of what is known about the regulation of Cldn expression in response to physiological or pathological stimuli, it was found that the expression of Cldn-14 (ion-restrictive Cldn) was increased in rat bone lining cells in response to chronic metabolic acidosis (29). In contrast, Cldn-15 and 16 (ion-permeable Cldns) were found to be downregulated, which would increase the barrier function of the bone membrane. Consequently, this will restrict mineral efflux from bone, and serve as a compensatory mechanism for slowing down bone loss in long standing acidemia (29). Moreover, Cldn-1 expression was found to be upregulated during wound healing, using an in vitro model (30).

Whereas it is well documented that osteoblast lineage cells express Cldns, virtually little is known about their expression profiles in bone resorbing osteoclasts. Nonetheless, we recently provided the first evidence that Cldns, specifically Cldn-18 is expressed in osteoclasts. Of the two tissue specific Cldn-18 isoforms, Cldn-18–1.1 (lung isoform) and Cldn-18–2.1 (stomach isoform), it was found that bone cells predominantly express the stomach isoform (32). However, it remains to be determined if other Cldns are expressed in osteoclast lineage cells and if the Cldns are important for the formation and activity of multinucleated osteoclasts.

Regulation of Cldn expression

Studies have shown that several growth factors, hormones, and cytokines play important roles in regulating Cldn expression and function. It has been demonstrated that TGF-β; regulates Cldn expression via Smad dependent and Smad independent pathways in several physiological and pathological states (9). Furthermore, several inflammatory cytokines (eg, IL-β, TNF-α, and IFN-γ) were implicated in modulating the expression of Cldns (9). For instance, it has been found that TNF-α downregulates barrier-forming Cldns (Cldn-1, 3, 4, 7, and 8) and upregulates pore-forming Cldns (eg, Cldn-2) in the intestinal epithelium, thereby increasing epithelial cell permeability (9). In addition, the expression of intestinal Cldn-2 and Cldn-12 was found to be upregulated by vitamin-D treatment, and has been implicated to play a role in the paracellular transport of calcium in the intestinal epithelium (33). Several transcription factors have been reported to regulate Cldn expression, with Snail and Slug being the most extensively studied. These transcription factors were shown to suppress the expression of various Cldns by direct binding to their gene promoters (25, 34, 35). For example, overexpressing Snail was found to down regulate the expression of Cldn-1, 2, 3, 4, and 7 (9, 35). Other transcription factors such as GATA-4, CDX-1, Grhl2, and SP1 have also been reported to regulate Cldns expression (9). The issue of whether these transcription factors are involved in mediating the transcriptional effects of growth factors and cytokines on Cldn expression in the target cells remains to be determined.

In terms of skeletal tissues, little is known about the physiological regulators of Cldn expression. A study by Hatakeyama et al (30), reported that Cldn-1, 2, and 6 expression in MC3T3-E1 mouse osteoblasts was upregulated by IGF-I, a key bone formation regulator. Despite the increased mRNA expression level of the three aforementioned Cldns, upregulation at the protein level was only observed for Cldn-1 (30). This study also demonstrated that the mechanism by which IGF-I upregulates Cldn-1 expression is mediated by the MAP-kinase pathway, as inhibition of this pathway diminished the increase in Cldn-1 expression (Figure 2A) (30). Since IGF-I is found to be upstream to Cldn-1, there is a possibility that the anabolic effect of IGF-1 on skeleton is mediated in part by Cldn-1 (30). Further studies involving knocking down of Cldn-1 in cells of osteoblast lineage are needed to evaluate whether IGF-I regulation of Cldn-1 expression is involved in regulating osteoblast formation and functions (Figure 2A).

Figure 2.

Proposed models for Cldns regulation and action in bone cells. A, Proposed model for Cldn-1 and Cldn-18 regulation/function in osteoblasts: IGF-I upregulates Cldn-1 expression via MAPK dependent pathway, while estrogen upregulates Cldn-18 in osteoblasts. This upregulation could be served as potential mechanisms for the osteoprotective effects of both IGF-I and estrogen. B, Proposed model for Cldn-18 action in osteoclasts: Cldn-18, via its PDZ binding motif, binds to the PDZ domain of ZO-2, to regulate the shuttling of ZO-2 from the cytoplasm to the nucleus where ZO-2 modulates the transcription of RANKL target genes. ZO-2: Zonula Occludens-2.

Recent work by our laboratory on bone cells revealed that Cldn-18 expression is regulated by estrogen (36). In fact, the mRNA levels of Cldn-18 were found to be reduced by 93% in bones of ovariectomized (estrogen deficient) mice compared with sham operated animals, whereas estrogen treatment increased Cldn-18 mRNA levels in bone cells, in vitro (36) (Figure 2A). Furthermore, Cldn-18 is thought to be a novel estrogen target in the skeleton, as its deletion protects from estrogen deficiency induced bone loss in mice (36).

Known functions of the Cldns

The role of individual Cldns has been investigated by employing loss or gain of function models in cells and whole animals (37). These studies showed that knockout (KO) of Cldns resulted in diverse phenotypes in different tissues as listed in (Table 2). For instance, targeted disruption of Cldn-1, 5, and 7 in mice resulted in postnatal lethality by affecting the skin epithelial barrier, blood-brain barrier, and kidney functions, respectively (38–40). By contrast, Cldn-16 overexpression in transgenic mice resulted in postnatal lethality, as a consequence of a defect in the skin epithelial barrier (41). Several renal function defects were observed in Cldn-2, 7, 16, and 19 KO mice. Also, the lack of Cldn-9, 11, and14 resulted in deafness (42–44). In addition, loss of Cldn-11 caused central nervous system nerve conduction abnormalities and male sterility (43, 45).

Table 2.

Phenotypic Changes Caused by Cldn Gene Deletions and Transgenics in Mice

Cldn	Phenotype	Ref
1	KO: Skin barrier defect resulting in dehydration and neonatal lethality	(38)
2	KO: Defect in leaky and cation selective barriers in kidney proximal tubule	(167)
5	KO: Blood-brain barrier defect and neonatal lethality	(39)
6	Overexpressing transgenic mice: skin barrier defect and neonatal lethality	(41)
7	KO: Growth retardation renal salt wasting, chronic dehydration, and neonatal lethality	(40)
9	Mutant: Deafness	(42)
11	KO: CNS myelin defect, male sterility, deafness	(43, 45)
14	KO: Deafness	(44)
15	KO: mega-intestine	(155)
16	KO: Renal divalent ion wasting resemble FHHNC in humans but without nephrocalcinosis	(168)
	Knock down: Renal calcium and magnesium wasting resemble FHHNC in humans	(157)
18.2	KO: Atrophic gastritis	(83)
	KO: Bone loss due to direct (noncanonical) effect on osteoclast differentiation.	(32, 85)
19	KO: Peripheral nervous system deficit resulted in behavioral changes and neuropathy	(163)
	Knock down: Renal calcium and magnesium wasting resemble FHHNC in humans	(169)
2 and 15	Double KO: Malnutrition due to decrease nutrient absorption resulting in neonatal lethality	(105)
11 and 14	Double KO: CNS myelin defect, male sterility, and deafness	(46)

CNS indicates central nervous system, FHHNC indicates familial hypomagnesemia with hypercalciuria and nephrocalcinosis.

To understand the interaction between Cldns, a double KO mouse approach has been employed. Specifically, the double Cldn-11/-14 KO phenotype was a combination of those observed in each of the single deletion animals, including male sterility, neurological deficits, and deafness. These findings suggest a lack of cooperation between these two Cldns (46). On the other hand, Cldn-2/-15 double deletion in mice resulted in infant death as a consequence of malnutrition (in a manner more dominant than in each of the single KOs), which highlights a cooperative interaction between these two Cldns in maintaining nutrient absorption in the intestine (47). In agreement with data from mouse models, the importance of Cldns in modulating various biological systems have been confirmed by linking mutations of Cldn genes to human diseases. Mutations in human Cldn-1 and Cldn-14 have been found in neonatal ichthoyosis and sclerosing cholangitis and nonsyndromic hearing loss, respectively (48–50). Several mutations in human Cldn-16 and -19 are associated with a hereditary renal disease known as familial hypomagnesemia with hypercalciuria and nephrocalcinosis (51, 52).

With regard to the skeleton, our studies have shown that mice with targeted disruption of Cldn-18 (stomach isoform) exhibited a 20–25% decrease in areal BMD of the total body, vertebra, and long bones (32). Although bone size was found not to be affected by Cldn-18 deficiency, μCT analysis revealed a 20% reduction in cortical thickness at the femoral middiaphysis in Cldn-18 KO mice. Additionally, Cldn-18 KO mice exhibited a dramatic decrease in trabecular bone volume, trabecular thickness, and trabecular number, whereas trabecular separation at both the L5 vertebra and distal femur metaphysis was increased. Histomorphometric analysis and in vitro assays of bone formation revealed that bone formation parameters were not affected by Cldn-18 deficiency. By contrast, Cldn-18 KO mice exhibited an 87% increase in tartrate resistant acid phosphatase (TRAP) labeled osteoclast surface in the trabecular bone, as well as an increase in osteoclast number. The increased bone resorption observed in Cldn-18 KO was also confirmed by measuring the serum and mRNA levels of osteoclastogenic marker genes. In addition, the number of TRAP positive multinucleated cells was greater in bone marrow macrophage cultures derived from Cldn-18 KO mice that were induced to differentiate in the presence of receptor activator of nuclear factor kappa B ligand (RANKL) and macrophage colony stimulating factor, in vitro. These data clearly demonstrate that the reduced BMD reported in Cldn-18 mice was due to increased osteoclast formation and bone resorption that highlights Cldn-18 as a novel negative regulator of bone resorption.

Because Cldn-18 deficiency in KO mice induced bone loss via a RANKL-mediated increase in bone resorption, an interaction between Cldn-18 deficiency and conditions that increase bone resorption via RANKL signaling, eg, hypocalcaemia and ovariectomy, has been predicted (53, 54). To evaluate this possibility, calcium deficiency was achieved by subjecting Cldn-18 KO and control mice to either a normal calcium (0.6%) or low calcium (<0.01%) diet for 2 weeks and the skeletal phenotype was subsequently evaluated by DXA and ìCT analyses (32). These studies revealed that bone resorption parameters were significantly increased by the combined deficiency of calcium and Cldn-18 compared with either calcium or Cldn-18 deficiency alone (32), but the underlying mechanisms that contributed to the interaction between the signaling pathways induced by calcium deficiency and Cldn-18 deficiency remained to be established. By contrast and unexpectedly, evaluation of the skeletal phenotype of ovariectomized or sham operated Cldn-18 KO mice revealed that ovariectomy failed to induce a significant change in BMD and trabecular architecture at different skeletal sites (36). Based on these findings and given that its expression is regulated by estrogen, Cldn-18 may act downstream of estrogen and mediate (at least in part) its effects on osteoclasts (36). It is worth noting that several line of evidence have suggested that the osteoprotective effect of estrogen is in part mediated by modulating osteoblast expression of cytokines that regulate osteoclast activity (55, 56). Given the fact that Cldn-18 is globally disrupted in both osteoblasts and osteoclasts, there is a possibility that estrogen regulation of Cldn-18 expression in osteoblasts is mediating the beneficial effect of estrogen on skeleton (Figure 2A). Thus, conditional disruption of Cldn-18 in different bone cells will be more informative.

Interestingly, humans with a Cldn-14 variant demonstrated a lower BMD in the spine and the hip than their normal counterparts, which provides further evidence that Cldns play an important function in bone (57). Taken together, it is clear that although much is known regarding the role of Cldn-18, the skeletal phenotype of other Cldns and their contribution to the genesis of osteoporosis remains to be investigated.

Molecular mechanisms of Cldn action

Cldns are multifunctional proteins. They regulate paracellular transport of ions, solutes, and water, and serve as a fence that divides apical and basolateral domains of plasma membranes, known as the “canonical function”. Aside from functioning as tight junctions, interestingly, Cldns have also been shown to participate in intracellular signaling that controls cell proliferation and differentiation, known as a “noncanonical function”.

Cldns are the major determinant of paracellular permeability in epithelial and endothelial cells, acting (canonically) as barriers or pores to decrease or increase permeability, respectively. Several in vitro studies have demonstrated that both barrier and pore forming Cldns are size and charge selective (Table 1) (9). In MDCK and LLC-PK1 cell lines, Cldn-2, 10b, 15, and 16, selectively increased cation permeability in tight junctions, whereas Cldn-1, 4, 5, 6, 8, 11, and 14 decreased cation permeability. Furthermore, Cldn-7, 10a, and 17 were found to serve as anion pores, but, in contrast, Cldn-7 and 19 act as anion barriers (9, 21). As mentioned before, the Cldn carboxy-terminus possesses several phosphorylation sites for various kinases including PKC, PKA, WNK, MLCK, MAPK, RhoK, and c-Src, which have been implicated for regulating tight junction assembly and function (58). In fact, several studies have reported that barrier/pore function of Cldns can be negatively or positively regulated by phosphorylation (19). It has been demonstrated that phosphorylation of Cldn-1 and Cldn-4 by PKC promoted their assembly into tight junctions and increased the barrier function (59–61). Conversely, dephosphorylation of Cldn-1 by PP2A resulted in their disassembly and decreased barrier function (62). As for negative regulation, Tanaka et al (63), reported that phosphorylation of Cldn-4 by EphA4 decreased its integration into tight junctions and increased paracellular permeability. It is to be noted though that the consequence of phosphorylation on the barrier function is rather complex and controversial. In fact, there are kinases that can phosphorylate distinct residues on the same Cldn with the outcome (positive or negative) depending on the external stimuli, and the physiological, and pathological conditions (64). Besides phosphorylation, it remains to be determined whether other post-translational modifications regulate canonical functions of Cldns.

The canonical function of Cldns in regulation of the paracellular transport of ions in bone has been suggested by several studies (29, 30, 65–67). There is substantial evidence that osteoblast and bone lining cells form an epithelial like bone membrane to control the paracellular ion exchange and maintain differential ion compositions between the plasma and bone extracellular fluid (8, 29, 30, 65–67). The expression and localization of certain Cldns in the bone lining cells support their function as barriers, a notion that was confirmed by measuring the transepithelial resistance of an osteoblast monolayer (8). As for regulation of the Cldn barrier function, more work must be done to address this issue. Nonetheless, in osteoblast like MC3T3-E1 cells, a reduction of paracellular permeability was observed as a consequence of MAPK activation by IGF-I (30). To this end, our previous study demonstrated that Cldn-18 disruption/overexpression did not influence paracellular transport of calcium ions in osteoclasts, thereby supporting a potential noncanonical function of Cldns in bone cells (32). Thus, some but not other Cldns may exert their canonical functions in bone cells.

Besides their canonical function as a barrier, gate, and fence, Cldns have recently started to emerge as mediators of cell signaling, eg, proliferation and differentiation (Table 1) (11). The noncanonical Cldn functions have been shown to involve interaction with adaptor proteins that shuttle between the plasma membrane and the nucleus, thereby regulating gene expression, cell proliferation, and differentiation (22). As mentioned earlier, Cldns have the capacity to interact with other PDZ domain containing cytoplasmic scaffolding proteins such ZO-1/2/3, via their carboxy- terminus PDZ-binding motif (12, 21). Although much remains to be investigated regarding the physiological significance of this interaction, research to date indicates that the interaction between Cldns and PDZ domain containing proteins is of importance in certain biological processes (68–71). Of the several adaptor proteins, ZO-1, the first tight junction protein identified, has been extensively studied (72). ZO-1 has three PDZ and SH3 domains that have been implicated in the regulation of cell proliferation. The negative effect of ZO-1 on cell proliferation is thought to be mediated by its SH3 domain interaction with the Y-BOX transcription factor ZONAB (73). The resultant cytoplasmic sequestration of ZONAB was found to reduce its interaction with important cell cycle regulators and cell cycle target genes (eg, CK4, and cyclin D) (73–76). Cldns have been also been implicated in regulating cell differentiation (77). For example, overexpression of Cldn-1 in intestinal epithelium was found to activate Notch-signaling, and in turn inhibit goblet cell differentiation (78). In our work, we found that Cldn-18 regulates RANKL-induced osteoclast differentiation via stimulating noncanonical signaling. We found that overexpression of Cldn-18 resulted in a dramatic inhibition of RANKL induced osteoclast differentiation in vitro, whereas bone marrow-derived macrophage precursors-derived from Cldn-18 KO mice formed fewer osteoclasts in the presence of M-CSF and RANKL, thus suggesting that Cldn-18 is a negative regulator of osteoclast differentiation. In terms of mechanisms for Cldn-18 action in osteoclasts, we found that overexpression of full-length Cldn-18 but not a mutant form of Cldn-18 with a deleted C-terminal PDZ binding motif had an effect on RANKL induced osteoclastgenesis, thus suggesting that the biological effects of Cldn-18 on osteoclasts are mediated by the PDZ binding motif. Of the several PDZ domain-binding proteins reported to directly bind to the C-teminal YV sequence of several Cldns, ZO-2 was found to be highly expressed and significantly upregulated by RANKL in osteoclasts. Consistent with this data, the Cldn-18 and ZO-2 interaction was confirmed by immunoprecipitation. Importantly, overexpression of Cldn-18 reduced ZO-2 nuclear translocation induced by RANKL, and this effect was abrogated with the deletion of the Cldn-18 C-terminal PDZ binding motif. Thus, we concluded that the negative effect of Cldn-18 on osteoclasts is mediated via sequestering ZO-2 in the membrane complex. Accordingly, we found that shRNA knockdown of ZO-2 inhibited RANKL-induced OC differentiation, in vitro. Consistent with this prediction, there is a significant body of evidence that ZO-2 has nuclear localization signals and that nuclear translocation is associated with increased gene expression (79–82). In this regard, ZO-2 knockdown in an osteoclast like cell line resulted in reduction of RANKL induced TRAP and cathepsin k expression, and inhibition of NF-κB and NFAT transcriptional activity. Collectively, we believe that the noncanonical effect of the loss of Cldn-18 in the regulation of RANKL-induced osteoclast differentiation is mediated by disruption of the interaction with ZO2, resulting in increased nuclear translocation of ZO2 (Figure 2B). In turn, this translocation increases the expression of important transcription factors involved in RANKL-induced osteoclast differentiation that ultimately leads to increased bone resorption.

Because the aforementioned findings were from mice in which Cldn-18 was globally disrupted, and since Cldn 18 is known to be expressed in other tissues, there is a possibility that the osteopenia phenotype is mediated in part by its disruption in tissues other than bone. To this end, Cldn-18 is the predominant form of Cldn in the stomach, and has been shown to play a major role in the physiology and pathology of the stomach epithelial barrier (83, 84). Recently, Hayashi et al (83), reported that targeted disruption of Cldn-18 in mice resulted in abnormalities in the gastric mucosa and atrophic gastritis via decreasing the paracellular barrier against H⁺ in the stomach epithelium. In agreement with their findings, we found that the loss of Cldn-18 negatively affected gastric acidity in adult mice (85). It is noteworthy that the role of gastric acidity in calcium absorption/metabolism has received more attention recently (86, 87). Interestingly, serum calcium levels were lower in Cldn-18 KO mice, whereas serum PTH levels were found to be elevated in Cldn-18 KO mice fed a normal calcium diet (32, 85). The possibility that decreased bone mass and increased bone resorption in Cldn-18 KO mice is a consequence of decreased calcium absorption due low gastric acidity, was examined by subjecting mice to either a normal (0.6%) or high calcium (2%) diet from birth until 10 weeks of age (85). Of note, correction of the serum calcium deficit with a high calcium diet did not correct the decreased BMD and increased bone resorption phenotype observed in Cldn-18 KO mice at different skeletal sites, thus ruling out the possibility that gastric abnormalities contributed to the osteopenia phenotype in these mice (85).

Based on the functions of Cldns known to date, it appears that Cldns do not follow a simple functional paradigm in various tissues. For example, in bone, the predicted role for Cldns involve maintenance of differential ion compositions between the plasma and bone extracellular fluid by regulating paracelluar transport of ions across bone lining cells (Table 3) (8, 29, 30). In addition, convincing evidence provided by our laboratory argues for a critical noncanonical role for Cldn-18 in regulating bone hemostasis (Table 3) (32, 85). Similarly, recent evidence has demonstrated that the expression of Cldn-2 and Cldn-12 was not correlated with calcium absorption under conditions of low calcium load (88). These data suggest that vitamin D mediated regulation of these Cldns has a noncanonical function in maintaining the integrity of intestinal epithelium (89). Furthermore, overexpression of certain Cldns has been implicated in certain types of cancer where these Cldns stimulate cell proliferation noncanonically (28). These findings open a broad range of avenues for research focused on understanding the emerging role of Cldns in bone cell signaling and biology. Our future understanding of the role and molecular mechanisms of Cldn action in bone could lead to the identification of novel targets for diagnosis and treatment of metabolic bone diseases including osteoporosis.

Table 3.

Known Actions of Cldns in Bone Cells

Cldn	Bone Cells Expression	Regulation of Expression	Known Actions
1	Osteoblast (8, 30)	Upreguated by IGF-I (in vitro) (30)	Reduction of paracelluar permeability by IGF-I (30)
		Upregulated during wound healing (in vitro) (30)
2 to 4	Osteoblast (8)	Not known	Not known
5	Osteoblast (8)	Not known	A potential role in regulating paracellular transport of ions across the bone lining cells (8, 29)
6 to 10	Osteoblast (8)	Not known	Not known
11	Osteoblast (8)	Not known	A potential role in regulating paracellular transport of ions across the bone lining cells (8, 29)
12	Osteoblast (8)	Not known	Not known
14	Osteoblast (8)	Upregulated in bone lining cells during chronic metabolic acidosis (in vivo) (29)	A potential role in regulating paracellular transport of ions across the bone lining cells (8, 29)
15 and 16	Osteoblast (8)	Downregulated in bone lining cells during chronic metabolic acidosis (in vivo) (29)	A potential role in regulating paracellular transport of ions across the bone lining cells (8, 29)
17	Osteoblast (8)	Not known	Not known
18	Osteoclast and osteoblast (8, 32)	Upregulated by estrogen (in vivo and in vitro) (36)	Negative regulator of RANKL induced osteoclast differentiation via sequestration of ZO-2 in the membrane complex (32, 85)
19, 20, 22, and 23	Osteoblast (8)	Not known	Not known

Future directions

While much is known regarding the expression patterns and function(s) of many Cldn family members in several tissues, little is known about their expression and role in bone. There is evidence that a number of Cldns are expressed in osteoblasts; however their expression during osteoblast differentiation is still not well defined. In addition, aside from Cldn-18, nothing is known about the expression of other Cldns in osteoclasts. Even though the skeletal phenotype of Cldn-18 is well established, that of other Cldns is yet to be examined. Thus, future research will therefore need to focus on the differential and comprehensive expression of Cldns in bone cells and both gain and loss of function experiments in vivo and in vitro to decode their role in skeletal development and maintenance.

Although the canonical Cldn functions have been well documented in epithelial and endothelial cells, more research is warranted to establish their role in bone. To this end, the canonical Cldn functions in controlling paracellular transport of ions across the bone lining cells has been suggested in the literature, but compelling evidence for this function is still so far lacking. This field is still in its infancy with many intriguing questions remaining unanswered: 1, What is/are the identity of ion restrictive/permeable Cldns in bone barriers? 2, Do they resemble the same functions documented in epithelial tissues? 3, Given the large number of Cldn proteins, how do these multiplicities influence their barrier function? 4, What is the interaction between these Cldns in regulating paracelluar permeability? 5, How is the barrier function regulated under both physiological and pathological states?

Beyond functioning as a tight junction, it is now clear that Cldns exert noncanonical functions by regulating cell signaling, which is becoming an emerging area of research. Whereas our recent work clearly documents a critical noncanonical role for Cldn-18 in regulating osteoclast differentiation in vitro, the following knowledge gaps for Cldn-18 and/or others warrant investigation: 1, Is Cldn-18 regulation of ZO-2 nuclear translocation involved in regulation of osteoclast differentiation in vivo, and if so, how does RANKL modulate ZO-2 release from the Cldn-18/ZO-2 complex at the plasma membrane? What are the upstream regulators of Cldn-18/ZO-2 expression and action in osteoclasts? 2, Does Cldn-18 as well as other Cldns interact with other proteins besides ZO-2 in osteoclasts? 3, What are the intracellular signaling pathways modulated by the Cldn-ZO-2 interaction? 4, Does Cldn-18 interact with other Cldns or tight junction proteins to regulate bone cell functions? 5, Does Cldn-18 exert different functions in bone versus other tissues (stomach and lung)?

Finally, an important issue that needs to be addressed is the involvement of Cldns in the pathogenesis of bone related diseases, such as osteoporosis. As mentioned earlier, a sequence variant of human Cldn-14 resulted in lower BMD in the spine and hip of affected individuals. However, if sequence polymorphisms in other Cldns contribute to variation in the skeletal phenotype remains to be evaluated. Thus, understanding how Cldns modulate osteoblast and osteoclast functions and overall bone homeostasis will provide new insights into the pathogenesis of osteoporosis and may highlight Cldns as novel targets for a therapeutic management of osteoporosis.