Regulation of tight junctions during the epithelium-mesenchyme transition:direct repression of the gene expression of claudins/occludin by Snail

The epithelium-mesenchyme transition (EMT) has attracted increasing interest from both developmental cell biologists and cancer researchers(Hay, 1995). EMT occurs in various steps in normal development, including mesoderm and neural-crest formation. Furthermore, the process of acquisition of an invasive phenotype by tumors of epithelial origin can be regarded as a pathological version of EMT. At the cellular level, EMT includes two distinct steps: decreased intercellular adhesion (to dissociate from the epithelial cellular sheets) and increased cell motility (to migrate into connective tissues). These steps are always associated with the total loss of epithelial cell polarity.

The tight junction (TJ) is an important structure that determines epithelial cell polarity and disappears during EMT. TJs constitute the epithelial junctional complex, together with adherens junctions (AJs) and desmosomes, and are located at the most apical part of the complex(Farquhar and Palade, 1963). TJs create the primary barrier to the diffusion of solutes through the paracellular pathway and maintain cell polarity as a boundary between the apical and basolateral plasma membrane domains(Schneeberger and Lynch, 1992; Gumbiner, 1993; Anderson and van Itallie, 1995; Tsukita et al., 2001). On ultrathin section electron microscopy, TJs appear as a series of discrete sites of apparent fusion, involving the outer leaflet of the plasma membranes of adjacent cells (Farquhar and Palade,1963). On freeze-fracture replica electron microscopy, TJs appear as a set of continuous, anastomosing intramembranous particle strands (TJ strands) (Staehelin,1974).

The molecular architecture of TJs has been unraveled rapidly in recent years. Three closely related PDZ-domain-containing proteins (ZO-1, ZO-2 and ZO-3) constitute the undercoat structure of TJs together with other peripheral membrane proteins such as cinglin, 7H6 antigen and symplekin(Stevenson et al., 1986; Gumbiner et al., 1991; Balda et al., 1993; Citi et al., 1988; Zhong et al., 1993; Keon et al., 1996; Mitic and Anderson, 1998; Tsukita and Furuse, 1999a). As constituents of TJ strands themselves, two distinct types of integral membrane proteins have been identified: occludin and claudins(Furuse et al., 1993; Furuse et al., 1998a; Furuse et al., 1998b). Both occludin and claudins bear four transmembrane domains but do not show any sequence similarity with each other. Claudins and occludin are thought to constitute the backbone of TJ strands and to modulate some functions of TJs,respectively (Tsukita et al.,2001). Claudins compose a multi-gene family consisting of more than 20 members (Morita et al.,1999; Tsukita and Furuse,1999b).

The question has naturally arisen of how TJs with such a complicated molecular architecture disappear during EMT. Recently, a zinc-finger transcription factor, Snail, has been implicated in the switching mechanism for EMT (Nieto, 2002). This gene was initially identified in Drosophila to be responsible for gastrulation (Grau et al.,1984; Nusslein-Volhard et al.,1984; Alberga et al.,1991). When chick embryos were treated with antisense oligonucleotides against Slug, the functional homolog of Snail in chicks, their mesoderm formation was affected(Nieto et al., 1994). Furthermore, Snail-deficient mice died at the gastrulation stage because of incomplete EMT (Carver et al.,2001). These mice developed a mesodermal layer that expressed some mesoderm-specific genes, but the mesoderm layer possessed some epithelial characteristics such as apical-basal polarity, retaining the expression of E-cadherin, a major cell adhesion molecule at AJs. This was consistent with our previous observation that a Drosophila Snail mutant failed to down-regulate E-cadherin expression at the ectoderm prior to gastrulation (Oda et al.,1998).

Recently, Snail has been shown to bind directly to E-boxes in the E-cadherin promoter and to repress E-cadherin expression directly, resulting in the destruction of AJs(Cano et al., 2000; Batlle et al., 2000). In this study, from the viewpoint of epithelial cell polarity, we examined the molecular mechanism for the destruction of TJs during EMT. We first established an in vitro mouse Snail-induced EMT using mouse cultured epithelial cells, and found that Snail directly suppressed the gene expression of claudins and occludin. Furthermore, we showed that E-boxes in the claudin and occludin promoters were responsible for this Snail-induced repression of their promoter activities. We believe that these findings provide a new insight into the molecular mechanism of EMT.

Rat anti-mouse occludin monoclonal antibody (mAb) (MOC37)(Saitou et al., 1997) and mouse anti-rat ZO-1 mAb (T8-754) (Itoh et al., 1993) were developed and have been characterized previously. Mouse anti-p120 mAb, mouse anti-cytokeratin 18 mAb (clone 18.04) and rabbit anti-claudin-3 pAb were purchased from Transduction Laboratories, Progen Biotechnik and Zymed Laboratories, respectively. Rat anti-mouse E-cadherin mAb(ECCD-2) was a gift from M. Takeichi (Kyoto University, Kyoto, Japan). Mouse Eph4 epithelial cells, mouse CSG1 epithelial cells, mouse NIH/3T3 fibroblasts,mouse L fibroblasts and human 293 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum.

The whole cell lysates of cultured cells were subjected to one-dimensional SDS-PAGE (12.5%), according to the method of Laemmli(Laemmli, 1970), and gels were stained with Coomassie brilliant blue R-250. For immunoblotting, proteins were electrophoretically transferred from gels onto nitrocellulose membranes, which were then incubated with the first antibody. Bound antibodies were detected with biotinylated second antibodies and streptavidin-conjugated alkaline phosphatase (Amersham, Arlington Heights, IL). Nitroblue tetrazolium and bromochloroindolyl phosphate were used as substrates for the detection of alkaline phosphatase.

Cells cultured on cover slips were rinsed twice with PBS and fixed with ice-cooled methanol for 10 minutes. After rinsing in PBS, the fixed cells were blocked with 1% bovine serum albumin (BSA) in PBS for 30 minutes and then incubated with primary antibodies for 30 minutes. They were then rinsed three times with PBS and incubated with appropriate secondary antibodies for 30 minutes. The secondary antibodies used were FITC-conjugated donkey anti-rabbit IgG polyclonal antibody (pAb), FITC-conjugated donkey anti-mouse IgG pAb and FITC-conjugated donkey anti-rat IgG pAb (Jackson Immunoresearch, West Grove,PA). After rinsing with PBS, the samples were embedded in 90% glycerol-PBS containing 0.1% paraphenylendiamine and 1%n-propylgalate.

Aliquots of total RNA (10 μg) were separated by 1.0% agaroseformaldehyde gel electrophoresis. Hybridization with digoxigenin (DIG)-labeled RNA probes was performed according to the manufacturer's protocol (Roche). Briefly, RNA was transferred onto positively-charged nylon membranes, followed by ultraviolet (UV) cross-linking. Nylon membranes were then hybridized with DIG-labeled RNA probes at 65°C in a buffer solution containing 50%formamide. After thorough washing and blocking, the membranes were incubated with alkaline-phosphatase-conjugated anti-DIG antibodies for 1 hour. After extensive rinsing, the membranes were incubated with the 1,2-dioxetane substrate CSPD (Tropix, Bedford, MA) and exposed to X-ray film. To obtain DIG-labeled probes, reverse-transcription PCR was performed. Total RNA was isolated according to the method developed previously(Chomczynski and Sacchi,1987).

Using a total cDNA population obtained from NIH/3T3 cells as a template,the full-length cDNA of mouse Snail was amplified by PCR and cloned into the pCAG vector (pCAG-mSnail). Similarly, a human Snail expression vector(pCAG-hSnail) was constructed after its full-length cDNA was amplified by PCR using a total cDNA library of SW480 cells as a template. A mouse Snail mutant lacking the SNAG (Snail/Gfi) domain was created according to the method described previously(Cano et al., 2000).

The cultured cells were transfected with one of the above expression vectors or an empty vector in serum-free DMEM containing 50 μM CaCl₂ using LipofectAmine Plus (Gibco BRL). After a 2-week selection in growth medium containing 400 μg ml^–1 of G418,resistant colonies were separated and then screened. Ten and six independent clones were established for Snail-expressing Eph4 and CSG1 cells,respectively.

Mouse claudin-3 (–1536 to +141), mouse claudin-4(–1669 to +105), mouse claudin-7 (–2336 to +195) and human occludin (–135 to +145) promoter fragments were cloned by screening the mouse and human genomic library with respective cDNA fragments,and were then inserted into the pGL3 vector (Promega). These reporter constructs (0.5 μg DNA well^–1) were transfected into cells cultured on 12-well dishes as described above and, 24 hours after transfection, firefly luciferase (Luc) and Renilla luciferase (RLluc)activities were measured using the Dual Luciferase Reporter Assay System(Promega) according to the manufacturer's instructions. RLluc activity was used to normalize Luc activity. In all experiments, the total amount of transfected DNA was standardized with an empty pCAG vector.

To mutate the E-box sequence in the mouse claudin-7 promoter, a Quickchange Site Directed Mutagenesis Kit (Stratagene) was used. The core sequence, 5′-CA(G/C)(G/C)TG-3′, was mutated to 5′-AA(G/C)(G/C)TA-3′.

The double-stranded oligonucleotides corresponding to the following E-box sequences were synthesized: the mouse claudin-7 E-box (+88 to +112),5′-GGTGCGCCGCACCTGCTCG-CCCGCA-3′, and the human occludinE-box (+14 to +42), 5′-CATCCGAGTTTCAGGTGAATTGGTCACC-3′. They were then end-labeled with ³²P-α-CTP using the Klenow enzyme. In mutated double-stranded oligonucleotides, the core sequence 5′-CA(G/C)(G/C)TG-3′ was mutated to 5′-AA(G/C)(G/C)TA-3′. The mRNA of hemagglutinin (HA)-tagged Snail was synthesized in vitro with T7 RNA polymerase using pCITE-HAmSnail as a template and then translated in rabbit reticulocyte lysate (Promega).

An electrophoretic mobility shift assay was performed essentially according to the method described previously (Kasai et al., 1992). Briefly, in vitro translated HA-Snail protein (or luciferase as a control) (1 μg protein) was incubated with ³²P-labeled oligonucleotides in 50 ml gel retardation buffer consisting of 12 mM HEPES (pH 7.8), 100 mM KCl, 15 mM ZnCl₂, 1 mM DTT, 12% (v/v) glycerol, 0.05% NP-40, 20 μg ml^–1 BSA, and 700 mg ml^–1 poly (dI-dC) for 30 minutes at room temperature. For the competition experiments, unlabeled oligonucleotides were added 10 minutes before the labeled ones. To detect the super-shift of the band, the solution was incubated with anti-HA mAb [or anti-green-fluorescent-protein(anti-GFP) mAb as a control] for 15 minutes at room temperature. Samples were then electrophoresed on 4% acrylamide gel with 0.5 M Tris-borate EDTA, and the gels were dried, followed by autoradiography.

DNA precipitations were carried out essentially according to the method described previously (Hata, 2000). Briefly, human 293 cells transiently expressing mouse HA-Snail were lysed, and the lysate was pre-absorbed using ImmunoPure streptavidin-agarose beads (Pierce) for 1 hour. The sample was then incubated with 1 μg of biotinylated double-stranded oligonucleotides corresponding to the E4 E-box sequence of the claudin-7 promoter, an E-box sequence of the occludin promoter, or their mutated sequences(see above), together with 10 μg of poly(dI-dC) for 16 hours. Biotinylated DNA-protein complexes were recovered using streptavidin-agarose beads for 1 hour, rinsed with HKMG buffer (10 mM HEPES, pH 7.9, 100 mM KCl, 5 mM MgCl₂, 10% glycerol, 1 mM DTT and 0.5% NP-40) and separated on SDS-polyacrylamide gels. Bound HA-Snail was detected by immunoblotting with anti-HA mAb.

To examine the effects of the Snail expression on epithelial morphology in cultured cells, dog MDCK II cells or human HT29M6 cells have previously been used (Cano et al., 2000; Batlle et al., 2000). In this study, we attempted to establish in vitro Snail-induced EMT using cultured mouse epithelial cells. For this purpose, we chose two mouse epithelial cell lines, Eph4 and CSG1, that show a typical cobblestone-like appearance under confluent culture conditions. When mouse Snail cDNA was introduced into these cells, they acquired a more fibroblastic phenotype(Fig. 1): they were rounded with many cellular protrusions, and their cell-cell adhesion appeared to be down-regulated. We isolated several independent stable clones for each cell line. Because all of the independent clones obtained showed the same phenotype, we mainly used one clone of the Eph4 transfectants expressing mouse Snail (Eph4-mSnail) for further analyses.

We examined the expression of AJ and TJ components in Eph4 and Eph4-mSnail cells by immunoblotting (Fig. 2A). Consistent with previous observations in dog and human epithelial cells (Cano et al.,2000; Batlle et al.,2000), the expression of E-cadherin in Eph4-mSnail cells was completely repressed. Furthermore, TJ integral membrane proteins such as claudin-3 and occludin became undetectable at the protein level. However, the undercoat proteins such as p120 (a cadherin-binding protein) and ZO-1 (an occludin/claudin-binding protein) did not alter their expression levels compared with the wild-type cells. This expression pattern of AJ and TJ components in Eph4-mSnail cells is very similar to that in NIH/3T3 fibroblasts. We then compared the subcellular distribution of these proteins between parental Eph4 and Eph4-mSnail cells by immunofluorescence microscopy(Fig. 2B). In Eph4-mSnail cells, E-cadherin became undetectable not only at cell-cell contact regions but also in the cytoplasm. In these cells, an AJ undercoat component, p120,showed no concentration at the cell-cell contact regions but was diffusely distributed in the cytoplasm. Interestingly, claudin-3 and occludin, which were concentrated at TJs in parental Eph4 cells, completely disappeared in Eph4-mSnail cells, whereas ZO-1 (a TJ undercoat protein) changed its localization from TJs to the cytoplasm. Non-junctional epithelial markers such as cytokeratin-18 appeared to be down-regulated by the Snail overexpression,but not completely.

The next question was whether the disappearance of occludin/claudins in Eph4-mSnail cells is due to direct repression of their transcription by Snail or to some indirect mechanism that facilitates their degradation. To address this question, we performed northern blotting(Fig. 3A). In Eph4-mSnail cells, the transcription of claudin-3, claudin-4, claudin7 and occludin, all of which were expressed abundantly in parental Eph4 cells, completely shut down together with the E-cadherintranscription. Consistent with the immunoblotting data, the mRNA levels of ZO-1 (and also p120; data not shown) did not alter significantly. Furthermore, the expression patterns of these mRNAs in Eph4-mSnail cells were very similar to those in NIH/3T3 fibroblasts and L fibroblasts(Fig. 3A,B). Taken together, we concluded that Snail directly and simultaneously represses the transcription of occludin and distinct species of claudins.

Random selection and transfection experiments identified a core of six bases [CA(G/C)(G/C)TG] as the consensus binding site for Snail(Muhin et al., 1993; Fuse et al., 1994; Inukai et al., 1999; Kataoka et al., 2000). This motif is identical to the E-box. Indeed, the human E-cadherinpromoter contains three E-boxes (Fig. 4) and Snail was reported to directly bind to these E-boxes to repress E-cadherin transcription(Cano et al., 2000; Batlle et al., 2000). Therefore, we next examined whether the gene transcription of TJ integral membrane proteins, claudins and occludin is also directly regulated by Snail. Then, we isolated the promoters of mouse claudin-3, claudin-4 and claudin-7. The putative transcription start point for each promoter was estimated according to the expressed sequence tag database(Fig. 4). Interestingly, these promoters contained six, eight and eight E-boxes, respectively. The human occludin promoter was isolated previously(Mankertz et al., 2000) and also contained one E-box (Fig. 4).

We then inserted the isolated fragments of the claudin-3,claudin-4 and claudin-7 promoters into the pGL3 plasmid upstream of the luciferase reporter gene, and transfected these reporter constructs into Eph4 epithelial cells and NIH/3T3 fibroblasts(Fig. 5A). In Eph4 cells, the claudin promoters induced a three- to tenfold increase in relative luciferase activity above that observed in NIH/3T3 cells, indicating that these promoter regions were sufficient to show the epithelium-specific activity. As previously reported (Mankertz et al., 2000), the isolated occludin promoter also showed similar epithelium-specific activity (data not shown). We next examined the ability of Snail to repress the claudin promoter activities. When the Snail expression vector and the claudin reporter constructs were co-transfected into Eph4 cells, the promoter activities of claudin-3,claudin-4 and claudin-7 were remarkably repressed(Fig. 5B), and this repression depended on the dose of Snail(Fig. 5C). Furthermore, when co-transfected with a Snail mutant lacking the N-terminal SNAG domain, which is required for the repressor activity of Snail in general(Grimes et al., 1996; Nakayama et al., 1998), the claudin-7 promoter activity was not repressed(Fig. 5D). Similar Snail-induced repression was also observed for the occludin promoter in human epithelial cells (HT29) (Fig. 5E). These findings indicated that the transcription of claudins and occludin was directly regulated by Snail by modulating the activities of their promoters.

To clarify further the molecular mechanism behind the Snail-induced repression of the claudin transcription, we examined a shorter fragment of the claudin-7 promoter in more detail(Fig. 6A). This short fragment,with five E-boxes, also showed epithelium-specific promoter activity (data not shown). We generated a series of reporter constructs that carried various combinations of mutated E-boxes and introduced them into Eph4 cells together with the Snail expression vector or an empty vector(Fig. 6B). Interestingly, when a single E-box was mutated, no significant impairment of Snail-induced repression was observed. However, as the number of mutated E-boxes was increased, the claudin-7 promoter became less sensitive to Snail(Fig. 6B,C), suggesting that E-boxes in the claudin-7 promoter are responsible for the Snail-induced repression.

We next performed an electrophoretic mobility shift assay. When the in vitro translated HA-Snail was incubated with ³²P-labeled double-stranded oligonucleotides corresponding to the sequence containing one of the E-boxes (E4) of the claudin-7 promoter(Fig. 6A), a single DNA-protein complex was observed (Fig. 7A). This complex was not formed with mutated oligonucleotides. Unlabeled wild-type oligonucleotides, but not mutated oligonucleotides, competed with the formation of the complex. Furthermore, anti-HA mAb, but not anti-GFP mAb,shifted the band of the complex upward. The same results were obtained when double-stranded oligonucleotides corresponding to the E-box in the occludin promoter were used (Fig. 7B). These findings indicated that Snail directly binds to the E-box in the claudin-7 and occludin promoters, at least in vitro.

Finally, to confirm the binding of Snail to the E-box (E4) of the claudin-7 promoters in the nuclear extract, we performed a biotinylated oligonucleotide precipitation assay(Fig. 7C, lanes 1-3). Biotin-labeled double-stranded wild-type or mutated oligonucleotides of the claudin-7 E4 sequence (Fig. 6A) were incubated with a nuclear extract prepared from 293 cells transfected with the HA-Snail expression vector or an empty vector. The biotinylated oligonucleotides were then recovered using streptavidin-conjugated agarose beads, and bound HA-Snail was detected by immunoblotting with anti-HA mAb. The HA-Snail in the nuclear extracts bound specifically to the wild-type, but not to the mutated, oligonucleotides(Fig. 7C). Similar specific binding was detected when the oligonucleotides of the occludin E-box sequence were used (Fig. 7C,lanes 4-6). Taken together, we concluded that Snail binds directly to E-boxes in the claudin and occludin promoters, and that Snail directly represses their activities.

The EMT is an important mechanism for the development of multicellular organisms, by which, for example, the mesoderm and neural crest are generated(Hay, 1995). Thus, a detailed description of EMT in molecular terms is an important issue in developmental and cellular biology. However, EMT occurs in spatially restricted regions in a transient manner during embryogenesis, which has made it technically difficult to examine EMT at the molecular level. Set against this situation, forward genetics in Drosophila has brought a breakthrough: Snail, a transcription factor, plays a key role in triggering EMT(Grau et al., 1984; Nusslein-Volhard et al., 1984; Alberga et al., 1991). Now, we can induce EMT in cultured epithelial cells simply by exogenously expressing Snail (Cano et al., 2000; Batlle et al., 2000). This in vitro system prompted a direct investigation into two questions. First, what types of signaling events upregulate the Snail expression to trigger EMT in a spatially and temporally regulated manner? Second, what genes are direct targets for the transcription factor, Snail?

The promoter of E-cadherin has been identified as a direct target for Snail. Snail directly represses E-cadherin promoter activity(Cano et al., 2000; Batlle et al., 2000). During EMT, the epithelial phenotypes of cells are converted into mesenchymal phenotypes: various gene products involved in the epithelial phenotypes must alter in their distribution and expression. Many of these alterations could be explained as secondary events induced by the down-regulation of E-cadherin. Indeed, several lines of evidence have shown that the disappearance or dysfunction of E-cadherin results in the loss of epithelial cell polarity(Takeichi, 1991; Rodriguez-Boulan and Nelson,1989). For example, when epithelial cells were cultured at low Ca²⁺ concentration (to cause E-cadherin dysfunction), the junctional complex, including TJs, was destroyed and epithelial cell polarity was lost. By contrast, it was reported that the Snail-induced phenotypic changes could not simply be attributed to the loss of E-cadherin. When E-cadherin was exogenously expressed in Snail-expressing epithelial cultured cells, which lost the expression of endogenous E-cadherin showing mesenchymal phenotypes, the epithelial phenotypes were not completely restored(Cano et al., 2000). Therefore,we must search for other direct targets for Snail for a better understanding of the molecular mechanism behind EMT.

In this study, we examined the behavior of claudins and occludin, major constituents of TJ strands, using the in vitro Snail-induced EMT system. Because TJs are the key structures responsible for establishing and maintaining epithelial cell polarity, claudins and occludin were expected to show a remarkable change in their distribution during EMT. Surprisingly,however, their expression completely shut down at the transcription level. Furthermore, we found that, similar to the Snail-based regulation of E-cadherin transcription (Cano et al., 2000; Batlle et al.,2000), Snail completely repressed the claudin and occludin promoter activities through its direct binding to E-boxes in these promoters. These findings unraveled the very sophisticated mechanism by which Snail caused EMT. Snail directly and simultaneously represses the expression of two distinct groups of important intercellular adhesion molecules, E-cadherin and claudins/occludin, which function at AJs and TJs,respectively.

The regulatory mechanism of the formation and destruction of TJs has been examined extensively, but promoter analyses of claudins and occludin are only just beginning. The expression of occludin was shown to be regulated by several factors at the transcription level(Mankertz et al., 2000; Chen et al., 2000; Li and Mrsny, 2000), and several transcription factors (such as the β-catenin/Tcf complex and Cdx homeodomain proteins/hepatocyte nuclear factor 1) were reported to bind directly to claudin-1 and claudin-2 promoters, respectively(Miwa et al., 2001; Sakaguchi et al., 2002). These findings indicated that the expression of claudins and occludin is finely controlled depending on the physiological and pathological conditions, but our data revealed that Snail eliminates these fine regulations to shut off the expression of distinct claudin species and occludin completely and simultaneously. It remains unclear why the expression of claudins and occludin must be repressed so completely in mesenchymal cells but, conversely, these data favored the notion that claudins and occludin, as well as E-cadherin, are key determinants of the epithelial phenotype, including epithelial cell polarity.

Finally, we should briefly discuss the relationship between Snail and TJs,in particular the expression of claudins, in malignant tumors. In some types of tumors of epithelial origin, Snail expression has been reported, when cancer cells acquired an invasive phenotype(Cano et al., 2000; Batlle et al., 2000). However,the expression patterns of claudins varied significantly depending on the type of tumor. In some breast cancers and squamous adenocarcinomas, claudins were frequently down-regulated (Kramer et al.,2000; Al Moustafa et al.,2002). It is thus possible that this repression is due to the upregulation of Snail expression, although this has not yet been examined in detail. Recently, E12/47 and SIP1 were also found to down-regulate E-cadherin expression through direct binding to single or paired E-boxes (Perez-Moreno et al.,2001; Comijn et al.,2001; Bolos et al.,2003). Therefore, the possibility could not be excluded that, in tumors, down-regulation of claudins occur via upregulation of these non-Snail-related silencers. In future studies, we must further clarify the pathological relevance of down-regulated claudin expression as well as the possible involvement of Snail in the alteration of claudinexpression in malignant tumors.

We thank Shu Narumiya for allowing us to use the luminometer. This study was supported in part by a Grant-in-Aid for Cancer Research and a Grant-in-Aid for Scientific Research (A) from the Ministry of Education, Science and Culture of Japan to S.T., and by JSPS Research for the Future Program to M.F.