A Single Gene Product, Claudin-1 or -2, Reconstitutes Tight Junction Strands and Recruits Occludin in Fibroblasts

Rabbit anti–mouse occludin polyclonal antibody (pAb) (F4) and rat anti– mouse occludin mAb (MOC37) were raised and characterized as described previously (Saitou et al., 1997; Sakakibara et al., 1997). Mouse anti-FLAG mAb, rabbit anti-hemagglutinin (HA) pAb, and mouse anti-T7 mAb were purchased from Eastman Kodak Co. (New Haven, CT), MBL Co. (Nagoya, Japan), and Novagen (Madison, WI), respectively. Mouse L cells and their transfectants were cultured in DME supplemented with 10% fetal calf serum (Nose et al., 1988).

Claudin-1 and -2 were tagged with FLAG-peptide or GFP at their COOH-termini. Expression vectors for FLAG–claudin-1 and FLAG– claudin-2 were previously reported as pCCL-1F and pCCL-2F, respectively (Furuse et al., 1998). To construct the expression vector of GFP– claudin-1 (pCCL-1G), the claudin-1–GFP fusion protein cDNA (Furuse et al., 1998) was introduced into pCAGGSneodelECoRI (Niwa et al., 1991), provided by J. Miyazaki (Osaka University, Osaka, Japan).

Occludin was expressed without a tag or with HA or T7 tag at its COOH terminus in L cells under the control of β-actin promoter. Expression vectors for occludin, HA-occludin, and T7-occludin (pBMOC, pBMOH1 and pBMOT, respectively) were constructed as follows. The full-length mouse occludin cDNA was cloned as described previously (Ando-Akatsuka et al., 1996). To construct a plasmid carrying HA-tagged mouse occludin cDNA (pMOH1), the DNA fragment encoding the HA peptide (YPYDVPDYA) was excised from pYT11 (Takita et al., 1995) by NheI digestion and subcloned into pBluescript SK(−) at the SpeI site, into which the mouse occludin cDNA with a BamHI site at the end of the open reading frame (ORF) was subcloned in frame using the BamHI site. Then a stop codon was introduced after the HA sequence by PCR. To obtain a plasmid carrying T7-tagged occludin cDNA (pMOT), the DNA fragment encoding T7 peptide (SMTGGQQMG) was excised from the pET3c vector (Novagen) by NheI and BamHI digestion. Four DNA fragments were linked in tandem and subcloned into pBluescript SK(−). A stop codon was introduced after the T7 sequence by SpeI digestion, into which mouse occludin cDNA with an EcoRI site at the end of the ORF was subcloned in frame using the EcoRI site. The expression vectors pBMOC, pBMOH1, and pBMOT were generated by subcloning of mouse occludin cDNA, HA-tagged mouse occludin cDNA excised from pMOH1 or T7-tagged occludin cDNA excised from pMOT into the expression vector pBATEM2 (Nose et al., 1988) from which E-cadherin cDNA had been removed.

Mouse L cells were used for transfection. 1-μg aliquots of each expression vector were introduced into L cells in 1 ml of DME using Lipofectamine Plus (GIBCO BRL, Gaithersburg, MD). Since pCCL-1F, pCCL-2F, and pCCL-1G each contained a G418 resistance gene, they were transfected singly. pBMOC was cotransfected with pSV2hph (hygromycin resistant), whereas pBMOT and pBMOH1 were cotransfected with pSTneoB (G418 resistant). After a 24- or 48-h incubation, cells were replated and cultured in DME containing 10% FCS and 500 μg of Geneticin (GIBCO BRL) or 200 μg of hygromycin to select stable transfectants. Geneticin- or hygromycin-resistant colonies were isolated after 10–14 d of selection. Clones expressing claudins or occludin were screened by fluorescence microscopy with antibodies or for green fluorescent protein (GFP) fluorescence. The stable transfectants of L cells expressing FLAG–claudin-1, FLAG–claudin-2, GFP–claudin-1, occludin, HA-occludin and T7-occludin were designated as C1FL, C2FL, C1GL, OL, OHL and OTL, respectively. L cells overexpressing both occludin and FLAG–claudin-1 (OC1FL cells) were obtained by transfection of OL cells with pCCL-1F by the same procedure as described above. We isolated several stable clones for each transfection experiment. Since clone 8 of C1FL, clone 6 of C2FL, clone 1 of C1GL, clone 1 of OL, clone 28 of OTL, clone 29 of OHL, and clone 18 of OC1FL expressed relatively large amounts of the introduced molecules, we used these clones for the demonstrations in this paper.

Subconfluent cells plated on coverslips were fixed with 1% formaldehyde in PBS for 10 min at room temperature. The fixed cells were then treated with 0.2% Triton X-100 in PBS for 10 min and washed three times with PBS. After soaking in PBS containing 1% BSA, the samples were treated with primary antibodies for 30 min in a moist chamber. They were then washed three times with PBS, followed by a 30-min incubation with secondary antibodies. FITC-conjugated goat anti–rat IgG antibody (Biosource, Camarillo, CA), rhodamine-conjugated goat anti–mouse IgG antibody (Chemicon, Temecula, CA), Cy3-conjugated goat anti–mouse IgG (Amersham-Pharmacia Biotech., Bucks, UK), and FITC-conjugated donkey anti–rabbit IgG (Chemicon) were used as secondary antibodies. Cells were washed three times with PBS and then mounted in 90% glycerol-PBS containing 0.1% para-phenylendiamine and 1% n-propylgatlate. Specimens were observed using a fluorescence Zeiss Axiophot photomicroscope (Carl Zeiss, Inc., Thornwood, NY).

One-dimensional SDS-PAGE (12.5%) was performed according to the method of Laemmli (1970), and proteins were electrophoretically transferred from gels onto polyvinylene difluoride membranes that were soaked in 5% skimmed milk followed by incubation with the first antibodies. Bound antibodies were detected with biotinylated second antibodies and streptavidin-conjugated alkaline phosphatase (Amersham Corp., Arlington Heights, IL). Nitroblue tetrazolium and bromochloroindolyl phosphate were used as substrates for detection of alkaline phosphatase.

For conventional freeze-fracture analysis, subconfluent L transfectants were fixed in 2% glutaraldehyde in 0.1 M sodium cacodylate buffer, pH 7.3, for 3 h at room temperature, washed with 0.1 M sodium cacodylate buffer three times, immersed in 30% glycerol in 0.1 M sodium cacodylate buffer for 2 h, and then frozen in liquid nitrogen. Frozen samples were fractured at −100°C and platinum-shadowed unidirectionally at an angle of 45° in Balzers Freeze Etching System (BAF060; Balzers Corp., Hudson, NH). The samples were then immersed in household bleach, and replicas floating off the samples were washed with distilled water. Replicas were picked up on formvar-filmed grids, and examined with a JEOL 1200EX electron microscope (JEOL, Peabody, MA) or a Hitachi H-700 electron microscope (Hitachi, Tokyo, Japan) at an acceleration voltage of 100 kV. The immunoelectron microscopic technique for examining freeze-fracture replicas was described in detail previously (Fujimoto, 1995). Subconfluent sheets of L transfectants were quickly frozen by contact with a pure copper block cooled with liquid helium gas (Heuser et al., 1979). Frozen samples were then fractured and shadowed as described above. The samples were immersed in sample lysis buffer containing 2.5% SDS, 10 mM Tris-HCl, and 0.6 M sucrose, pH 8.2, for 12 h at room temperature, then replicas floating off the samples were washed with PBS. Under these conditions, integral membrane proteins were captured by replicas, and their cytoplasmic domains were accessible to antibodies. The replicas were incubated with primary antibodies for 60 min, then with secondary antibodies coupled to 10- or 15-nm gold (Amersham Int.). The samples were washed with PBS and observed as described above.

Since Northern blotting revealed that cultured mouse L fibroblasts did not express any members of the claudin family (data not shown) or occludin (Saitou et al., 1997), L fibroblasts were used as parent cells for transfection in this study. After the FLAG tag was fused with the COOH termini of claudin-1 and-2, we introduced their cDNAs into cultured L fibroblasts, and L transfectants stably expressing molecules with the expected molecular masses were obtained (C1FL and C2FL cells, respectively) (Fig. 1,a). The subcellular distributions of these tagged proteins were then examined by immunofluorescence microscopy with anti-FLAG mAb. As shown in Fig. 2,a, b, d, and e, both introduced FLAG–claudin-1 and FLAG–claudin-2 were highly concentrated at cell contact sites as planes. Interestingly, close inspection revealed that in the cell adhesion sites both proteins were not diffusely distributed but concentrated in an elaborate network pattern (Fig. 2, c and f).

Next, to examine whether these FLAG-claudins were concentrated at cell contact sites through homophilic interactions, we further obtained stable L transfectants expressing claudin-1 tagged with GFP (C1GL cells). Similarly to FLAG–claudin-1, this GFP–claudin-1 was also concentrated at cell contact sites as planes, when C1GL cells were cultured under subconfluent conditions. As shown in Fig. 2, g and h, when C1GL cells were cocultured with C1FL cells, GFP–claudin-1 was co-concentrated with FLAG–claudin-1 at the cell adhesion sites as planes between adjacent C1GL and C1FL cells. Furthermore, both proteins showed the same network pattern of concentration in the cell adhesion sites (Fig. 2, i and j). Since claudins were not concentrated at the cell contact sites between adjacent C1FL (or C2FL) and parent L cells in mixed culture (data not shown), we concluded that these molecules were concentrated at cell adhesion sites through homophilic interactions.

C1FL cells were then examined by freeze-fracture methods after glutaraldehyde fixation. In parent L cells TJ strand-like structures were not detected, whereas in C1FL cells large networks of strands/grooves were frequently observed (Fig. 3,a). Judging from their size, these networks may correspond to the network patterns at cell contact planes observed by immunofluorescence microscopy (refer to Fig. 2, c, i, and j). In these networks, the strands were largely associated with the P face, and mostly continuous with intervening spaces of various widths (Fig. 3, a and b). On the E face, complementary continuous grooves were identified, containing only scattered small numbers of particles (Fig. 3, a and c). In Fig. 3 a, the fracture plane jumped from one membrane to another in a cell contact region, showing the E face of the nearer membrane and the P face of the further membrane with a narrow step-like discontinuity separating them. In these E and P faces, the strands and grooves appeared to constitute a continuous network, indicating that in these transfectants the extracellular space was completely obliterated at these strands/ grooves. These findings indicated that the claudin-1–induced network was similar to the TJ network in situ.

Similar TJ-like networks were observed also from the cell adhesion planes of glutaraldehyde-fixed C2FL cells by freeze-fracture electron microscopy (Fig. 3,d). Again, judging from their size, these networks may correspond to the network patterns at cell contact planes observed by immunofluorescence microscopy (refer to Fig. 2,f). In marked contrast to C1FL cells, in the claudin-2–induced networks TJ strands were fairly discontinuous at the P face leaving particle-free ridges between them (Fig. 3, d and e), and at the E face intramembranous particles formed chains that occupied the grooves (Fig. 3, d and f). Since the pattern of P face strands appeared to be continuous with that of E face grooves, the claudin-2–induced network is also concluded to function to obliterate the extracellular space (Fig. 3 d).

When C1FL and C2FL cells were examined by the freeze-fracture method without chemical fixation, the respective preferable associations of strand particles with P or E face became obscure (Fig. 4). Instead, both networks were characterized by ridges formed by aligned particles and short bars with intervening spaces on the P face. These P face strands from C1FL and C2FL cells were specifically labeled by anti-FLAG mAb, indicating that these claudins were directly involved in the formation of strand structures as major structural components (Fig. 4).

Next, we transfected mouse occludin cDNA into L fibroblasts, and obtained stable L transfectants expressing occludin which was detected as a 65-kD band by immunoblotting with anti-occludin mAb (OL cells) (refer to Fig. 1,b). The subcellular distribution of introduced occludin was then examined by immunofluorescence microscopy with anti-occludin mAb. As shown in Fig. 5,a, occludin was concentrated at cell contact sites in a punctate pattern. Again, to examine whether this concentration was due to the homophilic interaction of occludin, L transfectants expressing HA- or T7-tagged occludin were established (OHL and OTL cells, respectively). When these clones were co-cultured followed by double immunostaining with anti-HA pAb and anti-T7 mAb, HA- and T7-occludin were precisely coconcentrated in a punctate pattern between adjacent OTL and OHL cells (Fig. 5, b and c). We then concluded that similarly to claudins, occludin also showed a homophilic interaction resulting in concentration at cell contact sites.

In marked contrast to claudin-expressing cells, when glutaraldehyde-fixed OL cells were examined by the freeze-fracture method, large networks of strands/grooves were not detectable. Instead, small numbers of short and straight TJ strand-like structures (∼10-nm thick and up to ∼0.8 μm long) were observed (Fig. 5,d). These strands were largely associated with the P face (Fig. 5,d), and complementary grooves were found on the E face (Fig. 5,e). These structures were occasionally associated with gap junctions. Immunoreplica analyses with anti-occludin mAb for unfixed OL cells identified not only a small number of occludin-positive short strands (Fig. 5, f and g) but also a large number of occludin-positive dimple-like structures (Fig. 5, h–j). These dimple-like structures were too small to be interpreted in terms of P and E faces.

Since claudins and occludin are thought to be localized on TJ strands in situ, we finally cotransfected L fibroblasts with cDNAs encoding FLAG–claudin-1 and occludin. As shown in Fig. 1,b, stable L transfectants coexpressing both proteins were obtained (OC1FL cells), and the subcellular distributions of FLAG–claudin-1 and occludin were compared by doubly staining these cells with anti-FLAG mAb and anti-occludin mAb (Fig. 6, a–f). Similarly to C1FL cells, FLAG–claudin-1 was concentrated at regions of cell contact as planes in OC1FL cells (Fig. 6,a). Interestingly, in contrast to OL cells where occludin was concentrated at cell–cell borders in a punctate pattern (refer to Fig. 5,a), occludin in OC1FL cells was concentrated with FLAG– claudin-1 as planes at cell contact sites (Fig. 6,b). Furthermore, close inspection revealed that in the cell contact planes occludin was recruited to the claudin-1–positive network, although it was not evenly distributed along the claudin-1–positive network (Fig. 6, c–f).

Glutaraldehyde-fixed OC1FL cells were then examined by the freeze-fracture method. Large networks of strands were frequently observed, and significant differences were not discerned in the morphology of P face strands and E face grooves between OC1FL and C1FL cells (data not shown; refer to Fig. 3,a). When freeze replicas from quickly frozen OC1FL cells were double labeled with anti-FLAG mAb and anti-occludin pAb, both labels were admixed along strands (Fig. 6 g). These findings indicated that FLAG–claudin-1 and occludin were copolymerized into strands in OC1FL cells.

The structures of the elaborate TJ network revealed by the freeze-fracture technique has attracted increasing interest among cell biologists. In this study, we succeeded in reconstituting TJ-like network of strands/grooves in fibroblasts by introducing a single gene encoding claudin-1 or -2. Interestingly, the reconstituted networks were also detected by immunofluorescence microscopy. The claudin-induced network of strands/grooves was very similar to the in situ TJ networks in epithelial and endothelial cells (Staehelin, 1973, 1974): intramembranous particle strands and complementary grooves were observed on P and E faces, respectively. These strands and grooves occasionally branched and fused to form an anastomosing network. Furthermore, the fracture plane frequently jumped from one membrane (E face) to another (P face) maintaining continuity of the network pattern, indicating that apposed strands obliterate the extracellular space completely. Since the pattern of the reconstituted networks varied significantly from the cell–cell contact planes to planes (see the claudin-2–induced networks in Fig. 3,d and Fig. 4 b), it was difficult to quantitatively compare them with the in situ TJ networks.

Whether the TJ strands are predominantly lipidic in nature, i.e., cylindrical lipid micelles, or represent units of integral membrane proteins aggregated linearly has been the subject of some debate (Pinto da Silva and Kachar, 1982; Kachar and Reese, 1982; Verkleij, 1984). However, the detergent stability of TJ strands visualized by negative staining (Stevenson and Goodenough, 1984) and freeze-fracture (Stevenson et al., 1988) suggested that these elements were not composed solely of lipids. In this study, we showed that a single gene product, FLAG–claudin-1 or -2, induced TJ strand formation in fibroblasts, and that these strands were specifically labeled with anti-FLAG mAb by immunoreplica electron microscopy. As claudin-1 and -2 bear four transmembrane domains which allows their direct visualization as particles in freeze-fracture replicas, these results favor the notion that strands can be formed through polymerization of claudins within the plasma membrane. Furthermore, coculture experiments revealed that this polymerization was induced by homophilic interactions of claudins in adjacent cells. Of course, it is necessary to determine to what extent lipids contribute to the formation of strands in detail.

The most striking difference between claudin-1– and claudin-2–induced strands was found in the extent of the association of strand particles with the P face in glutaraldehyde-fixed samples: in L transfectants expressing FLAG–claudin-1, the strand particles were largely associated with the P face leaving continuous grooves occupied by a small number of particles in the E face. In contrast, in L transfectants expressing FLAG–claudin-2 the strands on the P face were fairly discontinuous, and in the E face particles formed chains that occupied the grooves. It has been shown that the extent of TJ strand particle association with the P face in glutaraldehyde-fixed samples varies depending on cell type as well as their developmental stage (for reviews see Staehelin, 1973, 1974; Pinto da Silva and Kachar, 1982). For example, in intestinal epithelial cells and secretory cells, freeze-fracture images of TJ strands/ grooves were similar to those of claudin-1–induced strands/ grooves in L cells, i.e., the P face-associated type of TJs (Friend and Gilula, 1972; Hull and Staehelin, 1976). On the other hand, in endothelial cells in nonneuronal tissues or in Sertoli cells in the testes, the images of TJ strands/ grooves were very similar to those of claudin-2–induced strands/grooves in L cells, i.e., the E face-associated type of TJs (Simionescu et al., 1975; Schneeberger, 1982; Nagano et al., 1982; Pelletier, 1988). These findings strongly suggested that the extent of TJ strand particle association with the P face is determined by the ratio of claudin-1 and -2 in TJ strands, or more correctly the combination ratio of claudin family members (Morita, K., M. Furuse, K. Fujimoto, and Sh. Tsukita, manuscript submitted for publication). As reported in situ, TJs (van Deurs and Luft, 1979; Pinto da Silva and Kachar, 1982), the extents of TJ strand association with the P face in L transfectants expressing claudin-1 and -2 were altered when they were observed without chemical fixation (refer to immunoreplica images in Fig. 4).

Recently, the extent of TJ particle association to the P face was demonstrated to be correlated well with the tightness of TJs in epithelial cells (Zampighi, G., R. Bacallao, L. Mandel, and M. Cereijido. 1991. J. Cell Biol. 115:479a) as well as endothelial cells (Wolburg et al., 1994; Kniesel et al., 1996). For example, cultured brain endothelial cells bore the E face-associated type of TJs and showed loose TJ barrier function, but when these cells were treated with astrocyte-conditioned medium to elevate their intracellular cAMP level the transendothelial electrical resistance was markedly increased with concomitant switching of TJs from the E face- to the P face-associated type (Wolburg et al., 1994). This correlation suggests that TJ strands consisting of claudin-1 are tighter than those consisting of claudin-2, and that the tightness of individual TJ strands in situ is determined by the combination of the members of claudin family in each TJ strand. This hypothesis will be evaluated by introducing the members of claudin family into cultured epithelial cells singly or in combination. Furthermore, the molecular mechanism behind the P face association of claudin-1 and behind the E face association of claudin-2 should also be clarified, paying special attention to the abilities of claudin-1 and -2 to interact with underlying cytoskeletal proteins.

In L transfectants expressing occludin (OL cells), occludin was concentrated at cell–cell contact sites in a punctate pattern through their homophilic interactions, but only a small number of short strands and grooves were formed on the P and E faces, respectively. Considering that overexpressed occludin formed short strands in multilamellar bodies in insect Sf9 cells (Furuse et al., 1996), occludin was concluded to also have a weak but significant ability to form TJ-like strands within the plasma membrane. Interestingly, these occludin-induced strands were occasionally associated with gap junctions. This is consistent with many previous observations that gap junctions were occasionally associated with TJs in situ (Elias and Friend, 1976; McGinley et al., 1977). Furthermore, in this study we have shown that occludin can be copolymerized into claudin-based strands in L transfectants (OC1FL cells), which is consistent with our previous observations that occludin, claudin-1, and -2 were colocalized on TJ strands in situ (Furuse et al., 1998). The question has naturally arisen as to the physiological functions of occludin. Occludin-deficient epithelial cells were shown to still bear well-developed and functional TJs (Saitou et al., 1998). Occludin-deficient mice were also born normally, but their growth rate was significantly reduced in comparison to controls (Saitou, M., K. Fujimoto, Y. Doi, M. Itoh, T. Fujimoto, M. Furuse, H. Takano, Sh. Tsukita, and T. Noda, manuscript in preparation). Taken together, it is likely that occludin is involved in modification of some function of claudin-based TJ strands. To clarify the functions of occludin, further transfection experiments with occludin and claudin in combination are required.

Previously we demonstrated that in liver at least three four-transmembrane domain proteins, occludin, claudin-1, and -2, constituted TJ strands. Then in the present study, we found that among these, claudin-1 and -2, but not occludin, showed a strong ability to form TJ strands within plasma membranes. These studies suggested a way to reconstitute TJs at the cellular level, which will be important to understand the molecular mechanism of barrier and fence functions of TJs as well as their regulation. The TJ structures induced in L cells by claudin-1 or -2 transfection were not zonula but puncta or fascia occludens, i.e., they did not surround individual cells continuously. Determination of what factors would be required to alter these punctae to zonulae occludens and to localize them at the most apical region of lateral membranes will be interesting.

We thank all the members of our laboratory (Department of Cell Biology, Faculty of Medicine, Kyoto University) for helpful discussions. Our thanks are also due to K. Yoshida (Department of Cell Biology, Faculty of Medicine, Kyoto University) for her excellent technical assistance.

This study was supported in part by a Grant-in-Aid for Cancer Research (06283225) and a Grant-in-Aid for Scientific Research (A) (8407006) from the Ministry of Education, Science and Culture of Japan.

TJ

tight junction

P face

protoplasmic face

E face

extracellular face

GFP

green fluorescent protein

HA

hemagglutinin

pAb

polyclonal antibody