Occludin-deficient Embryonic Stem Cells Can Differentiate into Polarized Epithelial Cells Bearing Tight Junctions

Rat anti–mouse occludin mAb (MOC37) and rabbit anti–mouse occludin polyclonal antibody (pAb) (F4) were raised against the cytoplasmic domain of mouse occludin produced in Escherichia coli (Saitou et al., 1997; Sakakibara et al., 1997). Mouse anti–rat ZO-1 mAb (T8-754) was raised and characterized as described (Itoh et al., 1993). Rat anti–mouse E-cadherin mAb (ECCD2) was provided by Dr. M. Takeichi (Kyoto University, Kyoto, Japan).

J1 ES cells (Li et al., 1992) and their mutant clones were cultured on mouse embryonic feeder cells (Rudnicki et al., 1992) in high glucose DME (GIBCO BRL, Gaithersburg, MD) supplemented with 20% FCS, 0.1 mM 2-mercaptoethanol (Sigma Chemical Co., St. Louis, MO), 1,000 U/ml leukemia inhibitory factor (LIF) (Amrad Co., Kew, Victoria, Australia), 0.1 mM nonessential amino acids (ICN Biomedicals Inc., Costa Mesa, CA), 3 mM adenosine, 3 mM cytosine, 3 mM guanosine, 3 mM uridine, and 1 mM thymidine (Sigma Chemical Co.). For embryoid body formation, 6 × 10⁶ ES cells were first cultured on gelatin-coated 10-cm culture dishes for 3 d in the same medium, and then subjected to suspension culture in three 10-cm bacterial dishes in the absence of LIF. The medium was changed every other day.

A λ phage 129/Sv mouse genomic library was screened using mouse occludin cDNA (nucleotides 1–424) as a probe, and then three overlapping clones were isolated. The overall structure of the mouse occludin genomic locus was determined by restriction and sequence analyses of these clones (see Fig. 1 A).

The targeting vector was designed to delete the entire sequence of exon 3 (273–945 nucleotides). A cassette consisting of SA (a splice acceptor sequence followed by triple splicing enhancers arranged in tandem [Friedrich and Soriano, 1991; Watakabe et al., 1993]), IRES (internal ribosomal entry sequence [Tsukiyama-Kohara et al., 1992]), LacZ (promoterless β-galactosidase sequence derived from pMC1871 [Pharmacia LKB Biotechnology, Uppsala, Sweden] that was preceded by a nuclear localization signal of SV-40 large T antigen [Kalderon et al., 1984]), loxP (Sternberg and Hamilton, 1981; Sauer and Henderson, 1988), pgk neo (Rudnicki et al., 1992), and loxP in this order was first constructed for positive selection. In this cassette, the pgk neo sequence was in the opposite orientation to the other sequences. A 7.5-kb BamHI fragment and a 1.8-kb HindIII– NheI fragment, which were located upstream and downstream from exon 3, were ligated upstream and downstream from the above cassette, respectively (see Fig. 1 A). A coding region of the diphtheria toxin A gene (Yagi et al., 1990) was then ligated downstream from the vector construct for negative selection against random integration of the vector.

The targeting vector was linearized at a unique NotI site located at the 5′ end of the 5′ homologous fragment, and then ES cells at passage 7–8 were electroporated with 20 μg of linearized targeting vector DNA using a Gene Pulser (Bio-Rad Laboratories, Hercules, CA) set at 400 V and 25 μF. Cells were plated on feeder cells in normal growth medium for 36–48 h, followed by selection with 175 μg/ml G418. Feeder cells were prepared from G418-resistant primary embryonic fibroblasts as described previously (Rudnicki et al., 1992). After 7–13 d, G418-resistant colonies were picked up and each colony was dissociated in Hepes-buffered saline (HBS) containing 0.1% trypsin and 1 mM EDTA, and then divided into two 96-well dishes. When each subclone became semiconfluent, cells in one well were frozen and those in the other well were used to isolate DNA for Southern blotting analysis (Laird et al., 1991). The colonies were screened individually by cleaving genomic DNA (15 μg) with XbaI and probing the Southern blots with 1-kb HindIII fragment just downstream from exon 4 (3′ probe). Correct targeting was confirmed by Southern blotting with an exon 2 probe (nucleotides 156–272) (5′ probe) (see Fig. 1 A). Targeted clones were also checked for single integration by hybridization with a neo probe.

To obtain occludin double knock-out ES cells, the single knock-out clones were plated at 6.0 × 10⁶ cells/10-cm dish and cultured for 10 d with an elevated concentration of G418 (20 mg/ml). 48 clones survived, and Southern blotting analysis with a 3′ probe indicated that these included two independent occludin double knock-out cell lines (clones 23 and 29).

Total RNA was isolated from ES cells or embryoid bodies using guanidine-thiocyanate and acid phenol/chloroform (Chomczymski and Sacchi, 1987) and cDNA was synthesized from 1 μg of total RNA using SuperScript II RNase H⁻ Reverse Transcriptase (RT; GIBCO BRL). Primers (upstream, 5′-TTGGGACAGAGGCTATGG-3′; downstream, 5′-ACCCACTCTTCAACATTGGG-3′) were designed to amplify a portion of occludin cDNA (nucleotide 487–1109). As a control for the presence of amplifiable RNA, hypoxanthine phosphoribosyl transferase primers were designed as previously described (Keller et al., 1993). PCR was performed in a 20-μl reaction volume containing 1 μg cDNA, 20 pmol of each primer, 0.25 mM of each dNTP, 1 unit of Taq polymerase (Boehringer Mannheim GmbH, Mannheim, Germany), and the buffer supplied by the manufacturer, under the following cycling conditions. For the occludin gene, an initial denaturation step of 3 min at 95°C was followed by 30 cycles of 1 min at 95°C, 1 min at 52°C, and 2 min at 72°C. The last cycle was followed by a further 10-min incubation at 72°C. Amplified PCR products were electrophoresed in 2% agarose gels and stained with ethidium bromide.

ES cells and embryoid bodies were homogenized in 2× SDS sample buffer (50 mM Tris-HCl, pH6.8, 2% SDS, 20% glycerol, 2% 2-mercaptoethanol, 0.01% bromophenol blue) on ice, sonicated, and boiled for 10 min. Total lysate (20 μg) was resolved by SDS-PAGE (10%) based on the method of Laemmli (1970). Proteins were then electrophoretically transferred from gels onto nitrocellulose membranes, followed by incubation with anti-occludin pAb (F4). For antibody detection, a blotting detection kit with biotinylated immunoglobulin and alkaline phosphatase-conjugated streptavidin (Amersham International Plc., Little Chalfont, UK) was used.

Embryoid bodies were frozen in liquid nitrogen, and sectioned on a cryostat at a thickness of 10 μm. Sections were mounted on glass slides, air-dried, fixed in 95% ethanol at 4°C for 30 min, and then in 100% acetone at room temperature for 1 min. After rinsing in PBS (150 mM NaCl, 10 mM phosphate buffer, pH 7.5) containing 1% BSA for 15 min, sections were incubated with a mixture of rat anti-occludin mAb (MOC37) and mouse anti–ZO-1 mAb (T8-754), or a mixture of rat anti–E-cadherin mAb (ECCD 2) and mouse anti–ZO-1 mAb (T8-754) for 1 h. They were then washed three times with PBS containing 1% BSA and 0.1% Triton X-100, and incubated with second antibodies for 30 min. As second antibodies, a mixture of FITC-conjugated goat anti–rat IgG (TAGO, Inc., Burlingame, CA) and rhodamine-conjugated goat anti–mouse IgG (Amersham International Plc.) was used. After washing in PBS, samples were embedded in 90% glycerol-PBS containing 0.1% para-phenylendiamine and 1% n-propylgalate, and examined using an MRC 1024 confocal fluorescence microscope (Bio-Rad Laboratories) equipped with a Zeiss Axiophot II photomicroscope (Carl Zeiss, Inc., Thornwood, NY).

Embryoid bodies were fixed with 2% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer, pH7.3, for 2 h at room temperature, followed by postfixation with 1% OsO₄ in the same buffer for 2 h on ice. Samples were stained en bloc with 0.2% uranyl acetate for 2 h at room temperature, dehydrated through a graded ethanol series, then embedded in Epon 812. Ultrathin sections were cut with a diamond knife, double stained with uranyl acetate and lead citrate, then examined using a 1200EX electron microscope (JEOL, Tokyo, Japan) at an accelerating voltage of 100 kV.

For conventional freeze fracture analysis, embryoid bodies were fixed in 2% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer, pH 7.3, for 2 h at room temperature, immersed in 30% glycerol, and then frozen in liquid nitrogen. Frozen samples were fractured at −110°C and platinum-shadowed unidirectionally at an angle of 45° in Balzer's Freeze Etching System (BAF 400T; 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 1200EX electron microscope (JEOL) at an accelerating voltage of 80 kV.

The immunoelectron microscopic technique for examining freeze fracture replicas was described in detail previously (Fujimoto, 1995). Small pieces of cystic embryoid bodies 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 domain was accessible to antibodies. The replicas were incubated with anti-occludin mAb MOC37 for 60 min, and then with goat anti–rat IgG coupled to 10-nm gold (Amersham International Plc.). The samples were washed with PBS and observed as described above.

Immunoelectron microscopy using ultrathin cryosections was performed essentially according to the method developed by Tokuyasu (1980; Fujimoto et al., 1992). Small pieces of embryoid bodies were fixed in 1% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4, for 1 h at room temperature. Samples were then infused with 2 M sucrose containing 20% polyvinylpyrrolidone at 4°C overnight, rapidly frozen in liquid nitrogen, and then cut into ultrathin sections at −110°C using a diamond knife with an FC-4E low temperature sectioning system (Reichert-Jung, Vienna, Austria). Cryosections were collected on carbon-coated Formvar-filmed grids, washed six times with PBS containing 10 mM glycine (PBS-G), and incubated with PBS-G containing 1% BSA for 10 min. Sections were again washed with PBS-G four times and then incubated with mouse anti– ZO-1 mAb (T8-754) for 1 h. After washing with PBS containing 0.1% BSA seven times, sections were incubated with goat anti–mouse IgG coupled to 10-nm gold (Amersham International Plc.) for 1 h. They were washed with PBS containing 0.1% BSA seven times, and with PBS six times. They were then fixed with 2% glutaraldehyde in 0.1 M phosphate buffer, pH7.4, for 10 min, and then washed with distilled water six times, followed by staining with 2% methylcellulose containing 0.5% uranyl acetate for 10 min. Samples were air-dried and examined in a 1200EX electron microscope (JEOL) at an accelerating voltage of 80 kV.

Tight junction permeability assay using surface biotinylation technique was performed as described by Chen et al. (1997) with some modifications. Fully expanded wild-type and occludin-deficient cystic embryoid bodies were incubated in freshly made 1 mg/ml NHS-LC-biotin (Pierce Chemical Co., Rockford, IL) in HBS(+) (HBS containing 1 mM CaCl₂ and 1 mM MgCl₂) for 30 min. In some experiments, wild-type cystic embryoid bodies were incubated with 1 mg/ml NHS-LC-biotin in HBS containing 10 mM EGTA for 30 min at room temperature. After washing with HBS(+) three times, the surface-labeled embryoid bodies were fixed with 1% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4, for 1 h at room temperature. Fixed samples were frozen in liquid nitrogen, and sectioned on a cryostat at a thickness of 12 μm. After rinsing in PBS containing 1% BSA for 1 h, sections were incubated for 1 h with a mixture of fluorescein-phalloidin (Pierce Chemical Co.) and XRITC-avidin (Pierce Chemical Co.), both of which were diluted 1:200 in PBS containing 1% BSA. The samples were embedded and examined by confocal fluorescence microscopy as described above.

Occludin genomic clones were obtained by screening a λ phage 129/Sv mouse genomic library using mouse occludin cDNA fragment as a probe. From overlapping clones, the genomic structure of the mouse occludin gene was partially clarified as shown in Fig. 1 A. The occludin gene contained at least five exons. The putative exon 2 contained the first ATG, exon 3 encoded the NH₂-terminal half of the occludin molecule from the first transmembrane domain to the second extracellular loop, and exon 4 encoded the fourth transmembrane domain and the initial part of the cytoplasmic tail. Considering that exon 5 was not found in the ∼30-kb genomic DNA fragment that we obtained, the occludin gene was expected to be >30 kb.

In the targeting vector, a cassette of SA IRES/LacZ/ loxP/pgk neo/loxP was flanked by upstream 7.5-kb and downstream 1.8-kb occludin DNA sequences (Fig. 1,A). The diphtheria toxin A gene driven by the MC1 promoter was ligated to the 3′ end of the construct to permit selection against random integration events. This targeting vector was designed with the expectation that homologous recombination between the vector and the occludin gene would result in deletion of exon 3, which consists of 673 bp (Fig. 1,A). The linearized targeting vector was introduced into ES cells, and cells were selected with G418. To screen for homologous recombination events, DNA from resistant clones were subjected to Southern blotting analysis with a probe corresponding to a sequence 3′ of the recombination site (3′ probe). The wild-type occludin allele displayed a 9.2-kb band on Southern blotting of XbaI- digested DNA, whereas the disrupted locus showed a 4.6-kb band (Fig. 1 B). Correct targeting was confirmed by Southern blotting with a 5′ probe, and targeted clones were also checked for single integration by hybridization with a neo probe. Of 200 G418-resistant clones examined, 9 had undergone a single homologous recombination event.

To prepare homozygous occludin-deficient cells (occludin double knock-out [DKO] cells), independent ES cell lines containing a single targeted occludin allele (6 × 10⁶ cells for each line) were cultured in the presence of 20 mg/ml G418. Southern blotting analysis showed that 2 of 48 ES clones resistant to high concentrations of G418 had undergone gene conversion, resulting in disruption of both alleles (Fig. 1,C). These two independent occludin DKO clones (clone 23 and clone 39) showed the same growth rate with undifferentiated morphology in the absence of G418 (see Fig. 2 A, e). Since these two DKO clones showed the same phenotype as far as was examined, the data obtained from clone 23 are mostly represented below.

As previously reported (Doetschman et al., 1985; Baribault and Oshima, 1991), when wild-type ES cells were subjected to suspension culture, cells aggregated to form “embryoid body” structures (EBs) (Fig. 2,A, a–d). Within 2–4 d in suspension culture, the loose cellular aggregates became compact to form “simple EBs,” delineated by an outermost layer of epithelial cells (visceral endoderm-like cells) (Fig. 2,A, b). In suspension culture for a further 5 d, these simple EBs gradually expanded to form “cystic EBs” (Fig. 2,A, c). The cystic EBs contained large cavities filled with secreted fluid (Fig. 2,A, d). We examined whether occludin DKO ES cells can form simple and cystic EBs. As shown in Fig. 2,A, f, occludin DKO ES cells formed typical simple EBs within 2–4 d. Considering that so-called Reichert's membrane was detected in their outermost layer, it was suggested that an epithelial layer was induced on their surface. Surprisingly, with the same time course as wild-type simple EBs, these simple EBs of occludin DKO ES cells started to expand to form typical cystic EBs, which contained large cavities filled with secreted fluid (Fig. 2 A, g and h). These findings suggested the sealing activity of outermost epithelial cell layers of occludin DKO EBs.

Next, we confirmed the loss of occludin expression in occludin DKO cells through the course of EB formation (ES cells, 4-d simple EBs, and 10-d cystic EBs) at both mRNA and protein levels (Fig. 2, B and C). Expression of occludin mRNA was examined by RT-PCR using total RNA to amplify an occludin cDNA fragment encoding the region from the first to the fourth transmembrane domain (Fig. 2,B). An occludin cDNA fragment of the expected size was amplified from wild-type ES cells and their simple/cystic EBs, whereas no bands were detected from occludin DKO ES cells and their EBs (Fig. 2,B). We then performed immunoblotting analyses of these ES cells and EBs using anti-occludin pAb. Consistent with the RT-PCR data, a trace amount of occludin was detected in wild-type ES cells, and its protein expression level increased gradually during the formation and development of EBs (Fig. 2 C). Of course, occludin was not detected at the protein level in occludin DKO ES cells or their EBs.

The loss of occludin expression in occludin DKO EBs was also confirmed by immunofluorescence microscopy (Fig. 3). Double immunofluorescence microscopy of frozen sections of simple EBs derived from wild-type ES cells revealed precise colocalization of occludin and ZO-1 at junctional regions of the outermost epithelial cell layer (Fig. 3, a–c). In contrast, no occludin signal was detected from occludin DKO EBs, whereas intense ZO-1 signals were still detected from junctional regions of the outermost epithelial cell layer (Fig. 3, d–f). The expression level and subcellular distribution of E-cadherin did not appear to be affected in occludin DKO EBs (see Fig. 6). In wild-type cystic EBs, occludin and ZO-1 continued to be precisely colocalized at junctional regions of outermost epithelial cell layers, although some structures (probably premature blood vessels) were ZO-1–positive but occludin-negative (Fig. 3, g–i). In occludin DKO cystic EBs, occludin disappeared from junctional regions of the outermost epithelial cell layers without affecting the expression and distribution of ZO-1 (Fig. 3, j–l). These findings suggest that occludin-deficient cells can differentiate into polarized epithelial cells, and that even in occludin-deficient epithelial cells the structural integrity of intercellular junctions is maintained. We then evaluated this speculation at the electron microscopic level.

We first compared the morphology of the outermost cell layers delineating occludin-deficient simple/cystic EBs with that of wild-type EBs by ultrathin section electron microscopy (Fig. 4). As far as was examined, there were no differences between wild-type and occludin-deficient EBs. Both wild-type and occludin-deficient simple EBs were delineated by polarized epithelial cells, where apical membranes with microvilli and basal membranes with basement membranes were easily identified (Fig. 4, a and b). Regardless of the occludin expression, these epithelial cells were further differentiated into highly polarized epithelial cells (Fig. 4, c and d). These cells were characterized by numerous apical microvilli as well as many secretory granules that were arranged in a polarized manner. Typical TJ structures with discrete sites of apparent membrane fusion were well developed at the most apical region of lateral membranes of wild-type as well as occludin-deficient epithelial cells (Fig. 4, e and f).

We then examined the TJ structures in cystic EBs by conventional freeze fracture electron microscopy (Fig. 5). As expected, continuous and anastomosing TJ strands were detected in freeze fracture replica images of outermost epithelial cell layers of wild-type EBs, although the number of TJ strands significantly varied from cell to cell and from EB to EB (Fig. 5, a and c). Again, no morphological differences were detected in freeze fracture replica images of TJ between wild-type and occludin-deficient cystic EBs. The number of TJ strands in occludin-deficient epithelial cells appeared to be distributed within a similar range to that in wild-type cells (Fig. 5, b and d), and the morphological appearance of P-face TJ strands and complementary E-face TJ grooves was also indistinguishable between wild-type and occludin-deficient epithelial cells (Fig. 5, e–h). Of course, by immuno-freeze replica analysis, the TJ strands of wild-type cells were heavily labeled with anti-occludin mAb, whereas those of occludin-deficient cells were not labeled (Fig. 5, i and j). These findings indicated that occludin-deficient ES cells can differentiate into polarized epithelial cells, where TJs are normally formed without occludin.

ZO-1 is known to be exclusively localized at TJs in epithelial cells expressing occludin, but to be localized in cadherin-based adherens junctions (AJs) in non-epithelial cells (Itoh et al., 1993, 1997). The existence of well-developed intercellular junctions in occludin-deficient epithelial cells led us to question whether the intense ZO-1 signal in these cells (see Fig. 3, e and k) was derived from TJ or AJ. Frozen sections of wild-type and occludin-deficient cystic EBs next were stained doubly with anti–E-cadherin and anti–ZO-1 mAbs, and observed by confocal microscopy. As shown in Fig. 6,A, the ZO-1 signal was detected more apically than the E-cadherin signal not only in wild-type but also in occludin-deficient epithelial cells. Considering that occludin and ZO-1 were precisely colocalized in wild-type epithelial cells, we concluded that ZO-1 still remained at TJ in occludin-deficient epithelial cells. Immunolocalization analysis using ultrathin cryosections confirmed the TJ-specific localization of ZO-1 in the occludin-deficient epithelial cells (Fig. 6 B).

Finally, we examined the barrier function of TJs in occludin-deficient epithelial cells delineating cystic EBs. The tracer experiment procedure developed by Chen et al. (1997) gave reliable and reproducible results. This procedure was originally developed for evaluation of the paracellular leakage of epithelial cells in Xenopus embryos. Wild-type and occludin-deficient cystic EBs were suspended in isotonic solution containing NHS-LC-biotin, which covalently cross-links to an accessible cell surface, for 30 min, washed, and then fixed with formaldehyde. Frozen sections of these fixed samples were incubated with XRITC-avidin to detect bound biotin, and observed by confocal microscopy. As shown in Fig. 7, both in wild-type and occludin-deficient EBs, only the apical surfaces of the outermost epithelial cell layers, but never their basolateral surfaces, were labeled with biotin. When EBs were incubated with NHS-LC-biotin in the presence of 10 mM EGTA, intercellular junctions were partly affected, then basolateral membranes of outermost epithelial cell layers were occasionally biotinylated. Furthermore, when the incubation time was prolonged up to 1 h, the same results were obtained (data not shown). These findings indicated that TJ lacking occludin can function as a primary barrier to paracellular leakage.

Recent technical improvements have enabled disruption of both alleles of any gene in ES cells by homologous recombination (Thomas and Capecchi, 1987; Mortensen et al., 1992). When EBs were formed from ES cells by suspension culture, ES cells differentiated into polarized epithelial cells (visceral endoderm-like cells) on their surface (Doetschman et al., 1985; Baribault and Oshima, 1991). Thus, by inducing EBs from ES cells homozygous for any gene mutation, we can directly assess the role of the gene product in epithelial cell differentiation. In this study, we adopted this strategy to clarify the function of occludin, which is the only known integral membrane protein localized at TJs. Southern blotting and immunoblotting analyses showed that both alleles of the occludin gene were successfully knocked out in two independent ES cell lines. Surprisingly, both occludin-deficient ES cell lines differentiated into polarized epithelial cells, which bore well-developed TJ. We failed to detect any morphological differences in the epithelial cells or TJ structures between wild-type and occludin-deficient EBs. We thus concluded that there are as yet unidentified TJ integral membrane protein(s) that can form TJ strands, although it is not clear whether such proteins exist as primary components or as compensatory components to safeguard epithelial function in the event of genetic mutations. Of course, it is also possible that occludin-deficient TJ strands are composed of inverted cylindrical micelles of lipids as has long been assumed (Kachar and Reese, 1982; Pinto da Silva and Kachar, 1982; Verkleij, 1984), but even in this case, there should be some proteinous components that stabilize the lipid micelles (Stevenson and Goodenough, 1984; Stevenson et al., 1988).

This conclusion is consistent with recent observations suggesting the existence of TJ lacking occludin. For example, the introduction of COOH-terminally truncated occludin into MDCK cells redistributed endogenous occludin to be concentrated in a discontinuous dotted manner along the cell–cell border, whereas the continuous network of TJ strands was not affected (Balda et al., 1996). Furthermore, when a synthetic peptide corresponding to the second extracellular loop of occludin was added to the culture medium, it drove out the endogenous occludin from junctional areas of cultured epithelial cells without affecting the gross epithelial cell morphology (Wong and Gumbiner, 1997). Also under physiological conditions, endothelial cells in non-neuronal tissues bore TJ but expressed only trace amounts of occludin (Hirase et al., 1997).

On the other hand, the above conclusion cannot be easily reconciled with other recent observations. In immuno-replica analyses, TJ strands were heavily labeled with anti-occludin antibodies (Fujimoto et al., 1995; Saitou et al., 1997), and the expression level of occludin in various cells is correlated well with the number of TJ strands (Saitou et al., 1997). Furthermore, overexpression of occludin resulted in an increase in the number of TJ strands in MDCK cells (McCarthy et al., 1996) and in the formation of short TJ strand-like structures in Sf9 cells (Furuse et al., 1996). These findings indicate that occludin is at least one of the major constituents of TJ strands. However, the present study showed that TJ strands can be formed without occludin. Considering that cystic EBs delineated by well-polarized epithelial cells were normally formed from occludin-deficient ES cells, the barrier and fence functions of TJ did not appear to be affected by the homozygous elimination of the occludin gene. As compared with cultured simple epithelial cells such as MDCK cells, EBs were not appropriate for the detailed analysis of the TJ barrier and fence functions, partly because various types of cells were differentiated randomly in EBs, and partly because the size/developmental stage of each EB and the uniformity/continuity of outermost epithelial cell layers were uncontrollable. Within this limitation, we found in this study that the occludin-deficient TJ functioned as a primary barrier to the diffusion of a small molecular mass tracer, NHS-LC-biotin. In good agreement, recently occludin-deficient mice were normally born, suggesting that the homozygous elimination of the occludin gene does not affect the TJ functions that are required at least for embryogenesis (Saitou, M., H. Takano, M. Itoh, M. Furuse, Sh. Tsukita, and T. Noda, manuscript in preparation). In contrast, as described in the Introduction, recent studies with COOH-terminally truncated occludin or a synthetic peptide of the second extracellular loop have indicated that occludin is directly involved in the barrier function of TJ (Balda et al., 1996; Chen et al., 1997; Wong and Gumbiner, 1997).

Of course, the most simple explanation for these discrepancies is that there are several distinct but similar occludin isotypes, which are functionally redundant in terms of TJ strand formation, i.e., these occludin isotypes can be co-oligomerized into TJ strands. Indeed, in the case of connexin, a four-transmembrane protein constituting gap junctions, >10 isotypes have been identified, which are expressed in tissues in various combinations to form gap junctions by homo- and/or hetero-oligomerization (Jiang and Goodenough, 1996; Kumar and Gilula, 1996; Brink et al., 1997). Effects of homozygous gene knock-out of connexin isotypes on gap junction formation or function was, however, reported to be restricted in some specific tissues (Reaume et al., 1995; Nelles et al., 1996; Anzini et al., 1997; Simon et al., 1997). However, the existence of so-called isotypes of occludin is unexpected. Occludin isotypes have not been identified despite intensive efforts by PCR using various primers, and no sequences similar to occludin were found in the data bases. It is thus premature to further discuss the relationship between occludin and other putative integral membrane protein(s) in TJ strands.

Occludin has also been implicated in the recruitment of ZO-1 to TJs through the direct binding of its cytoplasmic domain to the NH₂-terminal half of ZO-1 (Furuse et al., 1994). In non-epithelial cells lacking occludin expression, ZO-1 was reported to be directly associated with α catenin and to be concentrated at cadherin-based AJs (Itoh et al., 1997). We thus expected that in occludin-deficient epithelial cells ZO-1 might be concentrated in AJs. Unexpectedly, however, ZO-1 was still localized exclusively in the occludin-deficient TJs. These findings indicate that in addition to occludin there are some other binding partner(s) for ZO-1 in TJs. Considering that ZO-1 is a multi-domain protein belonging to the MAGUK family (Woods and Bryant, 1993; Anderson et al., 1995; Kim, 1995), it is likely that ZO-1 binds to more than two distinct integral membrane proteins in TJ strands. For example, at neuronal post-synaptic densities, PSD-95, another MAGUK family member, was reported to cross-link the cytoplasmic domains of three distinct transmembrane proteins: the NMDA receptor, potassium channel, and neuroligin (Kim et al., 1995; Kornau et al., 1995; Niethammer et al., 1996; Irie et al., 1997).

The present study clearly revealed that the molecular architecture of TJ strands is more complex than expected; i.e., that occludin is not the only integral membrane protein of TJs. To understand the function of occludin and to understand the molecular architecture of TJ, we should first identify other novel integral membrane protein component(s) of TJ strands. Identification of integral membrane proteins of TJ has been regarded as a difficult issue, but now we can use occludin as a probe to find such putative TJ components. Studies are currently underway along this line in our laboratory.

We thank all the members of our laboratory (Department of Cell Biology, Faculty of Medicine, Kyoto University) for helpful discussions.

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.

AJ

adherens junction

E-cadherin

epithelial cadherin

EB

embryoid body

ES

embryonic stem

DKO

double knock-out

HBS

Hepes-buffered saline

pAb

polyclonal antibody

RT

reverse transcriptase

TJ

tight junctions

ZO

zonula occludens