Tight junctions (TJs) seal the intercellular space of epithelial cells, while individual epithelial cells move against adjacent cells in cellular sheets. To observe TJs in live epithelial cellular sheets, green fluorescent protein (GFP) was fused to the N-terminus of claudin-3 (a major cell adhesion molecule of TJs), which was stably expressed at a level that was approximately 50% of that of endogenous claudin-3 in mouse Eph4 epithelial cells. Under confluent culture conditions, individual cells moved within cellular sheets, which was associated with the remodeling of TJs. However, during this remodeling, GFP-positive TJs did not lose their structural continuity. When TJs between two adjacent cells decreased in length during this remodeling, GFP-claudin-3 was frequently pinched off as a granular structure from GFP-positive TJs together with endogenous claudins. Co-culture experiments, as well as electron microscopy, revealed that the two apposed membranes of TJs were not detached, but co-endocytosed into one of the adjacent cells. Interestingly, other TJ components such as occludin, JAM and ZO-1 appeared to be dissociated from claudins before this endocytosis. The endocytosis of claudins was facilitated when the intercellular motility was upregulated by wounding the cellular sheets. These findings suggest that this peculiar internalization of claudins plays a crucial role in the remodeling of TJs, and that the fine regulation of this endocytosis is important for TJs to seal the intercellular space of epithelial cells that are moving against adjacent cells within cellular sheets.
In multicellular organisms, the internal environment must be isolated from the external environment, and further divided into various compositionally distinct fluid compartments. This isolation/compartmentalization is established by cellular sheets of epithelia and endothelia delineating the body surface and cavities. For these cellular sheets to function as barriers, there must be some seal to prevent diffusion of solutes through the intercellular route. Importantly, in the body, this intercellular sealing must be dynamically maintained (reviewed by Gumbiner, 1996; Schock and Perrimon, 2002). For example, the surface of the small intestine is lined by simple columnar epithelium, in which individual cells continuously move upward/towards the tips of villi as they mature (reviewed by Wilson et al., 1997): within these epithelial cellular sheets, their barrier function must be maintained, while individual cells are dynamically rearranged. Similar dynamic rearrangements of cells within epithelial cellular sheets, so-called `cell intercalation', have been examined in detail during the morphogenesis of invertebrates, such as germ band extension in Drosophila and notochord formation in ascidians (Irvine and Wieschaus, 1994; Munro and Odell, 2002).
In vertebrates, tight junctions (TJs) have been thought to be responsible for the intercellular sealing of simple epithelium (reviewed by Schneeberger and Lynch, 1992; Anderson and van Itallie, 1995; Balda and Matter, 1998; 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, 1965). On freeze-fracture electron microscopy, TJs appear as a set of continuous, anastomosing intramembranous particle strands (TJ strands) (reviewed by Staehelin, 1974). These morphological findings led to the following structural model for TJs: within the lipid bilayer of each membrane, the TJ strands, which are probably composed of linearly aggregated integral membrane proteins, form networks through their ramifications. TJ strands associate tightly with those in the apposing membrane of adjacent cells to form paired strands, where the intercellular distance becomes almost zero (reviewed by Gumbiner, 1987; Schneeberger and Lynch, 1992; Tsukita et al., 2001).
Three distinct types of integral membrane proteins have been shown to be localized at TJs; occludin, JAM and claudins (reviewed by Tsukita and Furuse, 1999; Tsukita et al., 2001; Gonzalez-Mariscal et al., 2003). Occludin, a ∼65-kDa integral membrane protein with four transmembrane domains, was identified as the first component of TJ strands (Furuse et al., 1993). However, several studies, including gene knockout analyses, have shown that TJ strands can be formed without occludin (Balda et al., 1996; Saitou et al., 1998). JAM with a single transmembrane domain was recently shown to associate laterally with TJ strands, while not constituting the strands per se (Martin-Padura et al., 1998; Itoh et al., 2001). By contrast, claudin is now thought to be a major constituent of TJ strands (Furuse et al., 1998a). Claudins with molecular masses of ∼23 kDa comprise a multi-gene family consisting of more than 20 members (Morita et al., 1999) (reviewed by Tsukita et al., 2001). Claudins also bear four transmembrane domains, but do not show any sequence similarity to occludin. When each claudin species, occludin or JAM was overexpressed in mouse L fibroblasts, claudin molecules, but neither occludin nor JAM, were polymerized within the plasma membranes to reconstitute paired TJ strands (Furuse et al., 1998b; Itoh et al., 2001). It was recently shown that heterogeneous claudin species are copolymerized to form individual TJ strands as heteropolymers, and that between adjacent TJ strands claudin molecules adhere with each other in both homotypic and heterotypic ways, except in some combinations (Furuse et al., 1999).
In recent years, the identification of claudins has enabled analysis of the molecular mechanisms behind the barrier function of simple epithelium in molecular terms, leading to the conclusion that claudin-based TJs are directly involved in the intercellular sealing (Sonoda et al., 1999; Simon et al., 1999; Wilcox et al., 2001). However, our knowledge is still fragmentary as to how this claudin-based intercellular sealing is dynamically maintained during active intercellular motility, i.e. between adjacent dynamically moving cells. In this study, as a first step to address this issue, we utilized GFP (green fluorescent protein) technology to visualize the dynamic behavior of TJs as well as claudins in cultured epithelial cells, and found a peculiar form of internalization of claudins, which was associated with the dynamic remodeling of TJs.
Materials and Methods
Antibodies and cells
Rabbit anti-GFP pAb, mouse anti-FLAG mAb and mouse anti-EEA1 mAb were purchased from Molecular Probes, Stratagene and Transduction Lab., respectively. Rabbit anti-claudin-3 pAb and mouse anti-claudin-4 mAb were purchased from Zymed Labs. Mouse anti-ZO-1 mAb (T8-754) and rat anti-occludin mAb (MOC37) were raised and characterized as described previously (Itoh et al., 1991; Saitou et al., 1997). Rat anti-E-cadherin mAb (ECCD2), rabbit anti-JAM pAb, mouse anti-LBPA mAb and rabbit anti-Rab7 pAb were generously donated by M. Takeichi (Riken, Kobe, Japan), T. Matsui (Kan Institute, Kyoto, Japan), T. Kobayashi (Riken, Saitama, Japan) and Y. Wada (Osaka University, Osaka, Japan), respectively. Epithelial cells, mouse Eph4, dog MDCK cells, were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum.
Expression constructs and transfection
The stop codon-deleted cDNAs of EGFP/ECFP/EYFP and stop codon-containing cDNA of CFP were amplified by PCR, and cloned into the pCAGGSneodelEcoRI vector (Niwa et al., 1991); pCAG-NGFP, pCAG-NCFP, pCAG-NYFP and pCAG-CCFP, respectively. To express claudin-3 tagged with EGFP at its N-terminus, claudin-1 tagged with ECFP or EYFP at its N-terminus, and ZO-1 tagged with CFP at its C-terminus, claudin cDNA with XhoI sites or mouse ZO-1 cDNA with EcoRI sites was inserted into pCAG-NGFP, pCAG-NCFP, pCAG-NYFP and pCAG-CCFP, respectively. These respective expression vectors were called pNGFP-mCld3, pNCFP-Cld1, pNYFP-Cld1 and pCCFP-ZO1.
The cultured epithelial cells were transfected with one of the above expression vectors in serum-free DMEM containing 50 μM CaCl2 using LipofectAmine Plus (GIBCO BRL). After a 2-week selection in growth media containing 400 μg/ml of G418, resistant colonies were separated and then screened by fluorescence microscopy.
Cells cultured on coverslips were fixed with 10% TCA on ice and incubated with 0.1% Triton X-100 for 10 minutes (Hayashi et al., 1999). For immunostaining with endosome-specific antibodies or detection of fluorescence signals of GFP-fusion proteins, samples were fixed with ice-cooled methanol for 10 minutes. After rinsing in PBS, the fixed cells were blocked with 1% BSA in PBS for 30 minutes, and then incubated with primary antibodies for 1 hour at room temperature. Cells were then washed with PBS several times and incubated with FITC- or Cy3-conjugated secondary antibodies (Chemicon International) for 1 hour. After rinsing with PBS, cells were mounted in Mowiol (Calbiochem).
To examine the membrane folding of EGFP-tagged claudin-3 in transfectants, some cells were first incubated with anti-GFP pAb or anti-claudin-3 pAb in the culture medium for 15 minutes on ice in the presence or absence of 0.1% Triton X-100 (Fig. 1D). Cells were then washed with PBS, fixed with 1% formaldehyde for 15 minutes, and incubated with 0.1% Triton X-100 for 10 minutes. These cells were processed for immunofluorescence microscopy as described above.
SDS-PAGE and immunoblotting
The whole cell lysates of cultured cells were subjected to one-dimensional SDS-PAGE (15%). For immunoblotting, proteins were electrophoretically transferred from gels onto nitrocellulose membranes, which were then incubated with the primary antibody. Bound antibodies were detected with horseradish peroxidase-conjugated secondary antibodies (Amersham, Arlington Heights, IL). ECL plus reagents (Amersham, Arlington Heights, IL) were used as substrates for the detection of peroxidase.
Quantification of expression levels of endogenous and exogenous claudin-3
For quantification of expression levels of endogenous claudin-3 and exogenous N-terminally-fused GFP-claudin-3 (NGFP-Cld3) in Eph4 transfectants, we first expressed a C-terminal cytoplasmic domain of claudin-3 fused with GST at its C-terminus (Cld-3cyt-GST) in Escherichia coli, and purified it. The amount of this purified GST fusion protein was determined by Coomassie Plus Protein assay reagent (Pierce). We then compared the intensity of immunoblotted endogenous claudin-3 and exogenous NGFP-Cld3 bands in cell lysates of Eph4 transfectants with those of immunoblotted bands of various amounts of purified Cld-3cyt-GST. The intensity of the band was measured by densitometry using NIH image software.
Fluorescence microscopy and image analysis
For live observations, cells were cultured on glass-bottomed dishes for 24-48 hours. For wound healing experiments, cells cultured on glass-bottomed dishes at high density were wounded by scraping with an 18G needle. Images of cells were collected at 37°C with a DeltaVision optical sectioning microscope (Version 2.50; Applied Precision, Issaquah, WA) equipped with an Olympus IX70 (PlanApo 60/1.40 N.A. oil immersion objective) through a cooled charge-coupled device camera (Series300 CH350, Photometrics, Tucson, AZ) with appropriate binning of pixels, exposure time and time intervals. Time-lapse images collected every 60-90 seconds for 50-100 frames were accumulated. All images, including movies, presented in this study were taken as five optical sections along the z axis.
Ultrathin-section electron microscopy
Cells were cultured on Transwell filters (Corning) 24 mm in diameter for 48-72 hours, fixed with 2% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.3) for 2 hours at room temperature, and postfixed with 1% OsO4 in the same buffer for 2 hours on ice. Samples were stained en bloc with 0.2% uranyl acetate for 2 hours at room temperature, dehydrated through a graded ethanol series and then embedded in Epon 812. Ultrathin sections were cut with a diamond knife, double stained with uranyl acetate and lead citrate, then examined under a 1200EX electron microscope (JEOL, Tokyo, Japan) at an accelerating voltage of 100 kV.
Correct targeting of N-terminally fused GFP-claudin-3 to TJs
As cDNA constructs encoding mouse claudins were used in this study, we mainly used mouse epithelial cell lines, Eph4 cells, for live observation. As endogenous claudins, Eph4 cells primarily expressed claudin-1, -3, -4, -6 and -7 (data not shown). To construct GFP fusion proteins with claudins, we have so far fused GFP to the C-terminus of claudins (Furuse et al., 1998a; Furuse et al., 1998b; Sasaki et al., 2003). However, as a free C-terminus is required for direct binding to PDZ domains (Itoh et al., 1999), these C-terminally fused GFP-claudins were expected to lose their ability to bind to PDZ domain-containing proteins such as ZO-1 and ZO-2, major TJ scaffold proteins. Thus, in this study, we fused GFP to the N-terminus of claudin-3 (NGFP-Cld3; Fig. 1A), introduced its cDNA into Eph4 cells and obtained two independent clones of stable trasfectants (Eph4:NGFP-Cld3).
We first examined the degree of expression of NGFP-Cld3 relative to endogenous claudin-3 in these two stable clones (#1 and #2) in a quantitative manner as described in detail in Materials and Methods. As shown in Fig. 1B, the ratios of exogenous NGFP-Cld3/endogenous claudin-3 were calculated as 0.56 (#1) and 0.47 (#2). Therefore, we concluded that in both clones the expression level of NGFP-Cld3 was compatible to that of endogenous claudin-3 (Fig. 1B).
As shown in Fig. 1C, when Eph4:NGFP-Cld3 cells were cultured under confluent conditions and then immunostained with anti-claudin-4 mAb or anti-ZO-1 mAb, GFP signals of NGFP-Cld3 precisely coincided with those of endogenous claudin-4 and ZO-1 at TJs. Then, we confirmed the correct conformation of NGFP-Cld-3 in the membrane of Eph4:NGFP-Cld3 cells: cells were immunofluorescently stained with pAbs specific for GFP or the C-terminal cytoplasmic tail of claudin-3 in the presence or absence of Triton-X 100 (Fig. 1D). GFP fluorescence was detected from TJs irrespective of the detergent treatment, but immunofluorescence signals from GFP and the C-terminal tail of claudin-3 were detected only when cells were permeabilized with detergent. These findings favored the correct conformation of NGFP-Cld-3 in the membrane of Eph4:NGFP-Cld3 cells as depicted in Fig. 1A. Therefore, in this study, we used Eph4:NGFP-Cld3 cells for the live observation of the behavior of claudins and claudin-based TJs.
Dynamic remodeling of TJs during intercellular motility
Live observation by fluorescence microscopy revealed that even in the confluent culture of Eph4:NGFP-Cld3 cells, individual cells moved against adjacent cells while changing their shapes (Fig. 2A; Movie 1). During the process of this intercellular motility (90 minutes), NGFP-Cld3 continued to concentrate at cell-cell borders, i.e. TJs, without a loss of structural continuity. Furthermore, during this series of time-lapse observation, one cell divided into two daughter cells (Fig. 2A,B), but such a dynamic change of shape was not associated with the breaking of the continuity of NGFP-Cld3-containing TJs surrounding dividing cells. This observation was consistent with previous immunofluorescence and electron microscopic studies (Jinguji and Ishikawa, 1992; Baker and Garrod, 1993).
We then observed more closely the remodeling of NGFP-Cld3-containing TJs associated with intercellular motility (Fig. 3). This dynamic remodeling involved two basic processes; the elongation and shortening of individual TJs between two adjacent cells. During the elongation process, claudins including NGFP-Cld3 should be added to TJs, but under the conditions used in this study, we failed to obtain any information on how claudin molecules are incorporated into elongating TJs.
A peculiar internalization of claudins
By contrast, when the TJs between two adjacent cells decreased in length, GFP-positive granular structures were frequently observed to bud from GFP-positive TJs, which were pinched off and then translocated into the cytoplasm (Fig. 3; Movie 2). On the basis of the confocal microscopic analyses, these structures did not reside on the plasma membranes but were scattered in the cytoplasm. As claudins are integral membrane proteins, these observations suggested that, when TJs between two adjacent cells shorten, they are endocytosed as vesicular structures.
Considering that TJs are composed of two tightly apposed membranes, the question has naturally arisen as to how these membranes are endocytosed. As schematically depicted in Fig. 4, two possible pathways of endocytosis can be considered. In Model #1, the two apposed membranes are first detached, i.e. claudin-based cell adhesion is released, by an unknown mechanism, and then individual membranes are endocytosed into their own cells. In Model #2, the two apposed membranes are not detached, but co-endocytosed into one of the adjacent cells. To evaluate these two models, we co-cultured Eph4:NGFP-Cld3 cells and parental Eph4 cells under confluent conditions, and closely observed the borders between these two distinct types of cells (Fig. 5A,B). Interestingly, from GFP-positive TJs between these two types of cells, GFP-positive granular structures were frequently pinched off and translocated into the cytoplasm of not only Eph4:NGFP-Cld3 cells (data not shown) but also parental Eph4 cells (Fig. 5B; Movie 3). As GFP signals should be derived only from Eph4:NGFP-Cld3 cells, not from parental Eph4 cells, this finding clearly indicated that the TJ plasma membranes belonging to Eph4:NGFP-Cld3 cells are endocytosed into adjacent parental Eph4 cells. Therefore, we concluded that the scenario Model #2 in Fig. 4 is what actually occurs.
To check whether this model applies to other types of epithelial cells, we examined the behavior of claudins in MDCK cells, which endogenously express claudin-1, -2 and -4 (Furuse et al., 2001). We established MDCK cells transfectants exogenously expressing nontagged claudin-3 (MDCK-Cld3) or N-terminally FLAG-tagged claudin-1 (MDCK-FCld1), and co-cultured them. When these cells were double stained with anti-claudin-3 pAb (green) and anti-FLAG mAb (red), red and green granular structures were frequently observed in MDCK-Cld3 and MDCK-FCld1 cells, respectively (Fig. 5C,D). This finding indicated not only that a peculiar Model #2-like endocytosis of TJs occurs also in MDCK cells, but also that nontagged claudin (claudin-3) behaves similarly to the GFP- or FLAG-tagged claudin. Interestingly, a similar type of peculiar endocytosis of TJs was more frequently observed, when two adjoining epithelial cells were mechanically detached by their own motility under subconfluent culture conditions. In Fig. 6, two types of MDCK transfectants expressing CFP-claudin-1 or YFP-claudin-1 were co-cultured at low cell density. Under such conditions, when the two distinct types adhered with each other, `zigzag', not linear, TJs were formed. When these two cells were dissociated as a result of their own cell motility, the TJs appeared to be torn off onto each cell, and subsequently became granular structures. Most of these granules were double-positive for CFP and YFP, indicating that a Model #2-like endocytosis also occurs during the cell dissociation (Movie 4).
We then searched for the ultrastructural counterparts of these endocytosed vesicles of TJs using ultrathin-section electron microscopy. As shown in Fig. 7, close inspection identified characteristic double-membrane vesicles along TJs of cultured Eph4 cells. Irrespective of intensive efforts, immunolabeling for these structures with anti-claudin pAbs was not successful, partly because the number of such vesicles may not be very large and partly because these structures may not be stable during the immunolabeling process. However, they would be only one type of vesicle, the morphology of which is consistent with the behavior of endocytosed TJs (see Model #2 in Fig. 4).
The question of what happens to these internalized claudins over time has naturally arisen. We observed that the endocytosed claudin-containing vesicles did not appear to be recycled back to the plasma membranes or TJs. Therefore, we performed immunofluorescent staining of MDCK cells expressing GFP-claludin-4 using antibodies specific for various endosome markers, as these are antibodies that have been used for MDCK cells. As shown in Fig. 8, endocytosed claudin-containing vesicles were negative for early endosome markers such as EEA1 (early endosome antigen 1), but some of these vesicles were positive for late endosome markers such as LBPA (lysobisphosphatidic acid) (Kobayashi et al., 1998) and Rab7 (Zerial and McBride, 2001; Pfeffer, 2003). These findings suggested that claudins are internalized as endocytosed vesicles from TJs, which are finally converted to late endosomes.
Internalized NGFP-Cld3 and endogenous junctional proteins
The next question is whether endogenous claudins and other TJ proteins in Eph4:NGFP-Cld3 cells are co-endocytosed with NGFP-Cld3. First, Eph4:NGFP-Cld3 cells were immunostained with anti-claudin-4 mAb. As shown in Fig. 9A, in all GFP-positive vesicles, endogenous claudin-4 was detected, and conversely all the claudin-4-containing vesicles carried exogenously expressed NGFP-Cld3. Furthermore, when parental Eph4 cells were double stained with anti-claudin-3 pAb and anti-claudin-4 mAb, many claudin-3-positive vesicles were scattered around the cytoplasm, all of which were positive for claudin-4 (Fig. 9B). Therefore, we concluded that distinct species of endogenous claudins are co-endocytosed from TJs, and that this endocytosis is not artifactually induced by exogenously expressed NGFP-Cld3.
Very unexpectedly, however, other TJ components, occludin, ZO-1 (Fig. 9C) and JAM (Fig. 10) were not detected in these claudin-containing vesicles in parental Eph4 cells, suggesting that claudins are selectively segregated and internalized from TJs. Indeed, when TCA-fixed Eph4 cells were double stained with anti-occludin mAb/anti-claudin-3 pAb, anti-JAM pAb/anti-claudin-4 mAb or anti-ZO-1 mAb/anti-claudin-3 pAb, occludin/JAM/ZO-1 were co-distributed with claudins in tubular structures invaginated from TJs, but were absent in claudin-containing vesicles, which appeared to be pinched off from the tubular structures (Fig. 10).
To confirm the segregation of internalized claudins from other TJ components in live cells, we observed the dynamic behavior of exogenously expressed GFP-ZO-1 during the endocytosis of TJs in Eph4 cell. In good agreement with immunostaining in Fig. 9C, GFP-positive granules were hardly observed to bud off from GFP-positive TJs (data not shown). We then attempted to pursue the behavior of ZO-1 and claudin simultaneously in living cells: we observed Eph4 cell clones that co-express YFP-claudin-3 and CFP-ZO-1. As shown in Fig. 11A, both YFP-claudin-3 (green) and CFP-ZO-1 (red) were co-concentrated at TJs. Interestingly, time-lapse observation revealed that the endocytosed TJ granules contained YFP-claudin-3, but not CFP-ZO-1 (Fig. 11B; Movie 5). YFP-single positive granules budded off from YFP/CFP-double-positive TJs. These findings clearly showed in live cells that before the initiation of the endocytosis of claudins, claudins are segregated from other TJ components.
Regulation of the endocytosis of TJs
Finally, as a first step to understanding the regulatory mechanism for the endocytosis of TJs, we examined the relationship between the degree of intercellular motility and the frequency of the endocytosis of TJs. For this purpose, we performed the wound healing experiments: confluent cultures of Eph4:NGFP-Cld3 cells on coverslips were scratched manually with an 18G needle. After 2 hours culture, the dynamic behavior of GFP-positive TJs was observed in the third row of cells from the front of the wound (Fig. 12B), and it was compared to that of the nonwounded cell sheet. During the wound healing process, cells moved towards the wound actively, resulting in an increase in intercellular motility (Fig. 12A). Concomitantly, the frequency of the endocytosis of GFP-positive TJs appeared to be increased. Then, for quantification, we enumerated the GFP-positive granules in the cytoplasm of individual cells in the third row in the wounded sheet (Fig. 12B,C): in the nonwounded sheet, 14.2±6.14 granules (n=13) were found per cell, wheras 23.9±11.91 granules (n=11) per cell were detected in the wounded sheet (P<0.05). Therefore, we concluded that the increase in intercellular motility upregulates the endocytosis of TJs.
TJs seal the intercellular space between adjacent epithelial cells, i.e. create a primary barrier to the diffusion of fluid, electrolytes and macromolecules through the paracellular pathway. However, under various physiological conditions, individual cells continuously move against adjacent cells within epithelial cellular sheets (reviewed by Gumbiner, 1996; Schock and Perrimon, 2002). It appears difficult for these two events – intercellular sealing and motility – to occur simultaneously, but in epithelial cellular sheets TJs must overcome this difficulty. The question has naturally arisen as to what is the molecular mechanism behind this peculiar ability of TJs. As a first step toward answering this question, we began to observe the dynamic behavior of claudins, major cell adhesion molecules of TJs, using GFP technology. Recently, we successfully made time-lapse observations of the paired strands reconstituted from GFP-claudin-1 in L fibroblasts, and found that individual reconstituted paired claudin strands behaved more dynamically than ever expected (Sasaki et al., 2003). In the present study, we used GFP-claudin to pursue the dynamic behavior of TJs themselves more macroscopically in live epithelial cellular sheets.
Occludin was reported to be targeted to basolateral membranes (Matter and Balda, 1998) and incorporated into TJs with concomitant heavy phosphorylation (Sakakibara et al., 1997), but our knowledge on the molecular turnover of claudins is still fragmentary. Also, in the present study, live observation gave no information on how claudin molecules are supplied to TJs: no GFP-claudin-positive granules/vesicles were observed to be targeted into lateral membranes or TJs per se. The amount of GFP-claudins within individual transport vesicles would be too small to be visualized, or the number of transport vesicles carrying claudins is fairly small under the confluent culture conditions.
However, we succeeded for the first time in capturing the process of endocytosis during the shortening phase of TJs between two adjacent epithelial cells. Live observations with the co-culture system as well as ultrathin-section electron microscopy revealed that the mode of this endocytosis is peculiar. Without being detached, the tightly apposed membranes of TJs were endocytosed together into one of the adjoining cells. This means that half of the plasma membranes/strands contributing to TJ structures are engulfed by adjacent cells. These engulfed vesicles carrying claudins would finally be integrated into the conventional endosome pathway (Fig. 8). In earlier studies, freeze-fracture replica electron microscopy showed that under various conditions that would facilitate the degradation of TJs, some intracellular vesicles carried TJ strands (Staehelin, 1973; Polak-Charcon and Ben-Shaul, 1979; Madara, 1990), leading to the speculation that TJ structures are endocytosed, but a peculiar form of endocytosis, like Model #2 in Fig. 4, had not been expected. This type of endocytosis was more frequently observed, when two adjoining cells, between which TJs had been established, were detached from each other by their own motility under nonconfluent conditions (Fig. 6).
Of course, in this type of live cell observation using GFP technology, we must carefully evaluate the impact of overexpression and GFP-tagging of the exogenously expressed molecules on findings. First, according to quantitative analyses (Fig. 1B), the expression level of NGFP-Cld3 was ∼0.5-fold that of endogenous claudin-3, indicating that exogenous claudin-3 was not `overexpressed'. Second, the subcellular localization of NGFP-Cld3 was very similar to that of endogenous claudins in fixed parental Eph4 cells at the immunofluorescence microscopic level, and all the NGFP-Cld3-positve vesicles carried endogenous claudin-4 (Fig. 9A,B). Furthermore, the peculiar `eat-each-other' endocytosis of claudins was observed in parental Eph4 cells adjoining Eph4:NGFP-Cld3 cells (Fig. 5A,B) as well as in MDCK cells expressing nontagged claudin-3 adjoining MDCK cells expressing FLAG-tagged claudin-1. Thus, it is reasonable to consider that the behavior of NGFP-Cld3 in Eph4 cells can be regarded as that of endogenous claudins. Importantly, during the course of the evaluation of overexpression and tagging of NGFP-Cld3, we noted an unexpected fact: during the internalization of claudins from TJs, claudins are segregated from other TJ components such as occludin, JAM and ZO-1 to generate claudin-enriched vesicles lacking occludin/JAM/ZO-1 (Fig. 9C and Fig. 10). At present, the molecular mechanism as well as the physiological relevance of this segregation of claudins remains totally elusive. For a better understanding of not only TJ turnover itself but also epithelial polarization, this issue should be analyzed in further detail in the future.
To date, many efforts have been made to clarify how intercellular junctions/adhesion molecules are degraded in the process of physiological turnover in general (Siliciano and Goodenough, 1988; Kartenbeck et al., 1991; Harhaj et al., 2002; Ivanov et al., 2003). When calcium ions were depleted from the culture medium, the apposed membranes of cadherin-based adherens junctions (AJ) and desmoglein/desmocolin-based desmosomes (DS) dissociated, and individual membranes were endocytosed into their own cytoplasm (Mattey and Garrod, 1986; Kartenbeck et al., 1991). Interestingly, recent studies reported that the apposed membranes of connexin-based gap junctions (GJ) are also endocytosed together into one of the adjoining cells without being dissociated (Jordan et al., 2001). Considering that both claudins and connexins bear four transmembrane domains, it would be interesting to speculate that this `eat-each-other' type of endocytosis is characteristic for TJ and GJ, but whether this type of endocytosis occurs also in AJ and DS should be examined in the future.
As a next step, we should examine how the endocytosis of TJs is regulated under physiological conditions. In this study, we showed that the upregulation of intercellular motility facilitated the endocytosis of TJs. Interestingly, as discussed above, we found that, when TJs were endocytosed, claudins appeared to be dissociated from other TJ components. These findings led us to the following hypothesis. First, some machinery detects the increase in intercellular motility; second, this information is transmitted to TJs locally; third, this signal dissociates claudins from other TJ components including underlying cytoskeletons; and fourth, the endocytosis of claudins is facilitated. It is still premature to discuss further the regulatory mechanism for the endocytosis of claudins, but through a very sophisticated mechanism, TJs would be able to seal the intercellular spaces while permitting active intercellular motility in epithelial cellular sheets.
Movies available online
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 to S.T., and by JSPS Research for the Future Program to M.F.