PATJ connects and stabilizes apical and lateral components of tight junctions in human intestinal cells

Formation of a polarized epithelial layer is a fundamental step during the development of multicellular organisms. This process involves the coordinated action of adhesion molecules, actin remodeling and spatial organization of membrane traffic (for reviews, see Spiliotis and Nelson, 2003; Zahraoui et al., 2000). Several polarity protein complexes are known to act in a spatio-temporal manner to build the apical and lateral domains of epithelial cells in Drosophila. These complexes are the Par3/6/aPKC complex and the Crumbs/Stardust/DPATJ complex on the apical side and the Scribble/Discs large/Lethal giant larvae complex, on the lateral side (for a review, see Gibson and Perrimon, 2003). The balance between the activities of these complexes regulates the formation and the stabilization of a continuous belt of adherens junctions localized at the border between the apical and lateral domains (Bilder et al., 2003).

Crumbs is an apical transmembrane protein that binds to Stardust (Sdt), a cortical membrane-associated guanylate kinase (MAGUK) and both are synthesized at the beginning of gastrulation (Bachmann et al., 2001; Hong et al., 2001; Knust et al., 1993). Sdt binds to DPATJ, which is expressed maternally before Crb and Sdt (Bhat et al., 1999). However, as no mutant exists, there is no description of a loss-of-function phenotype in flies. Therefore, its role in the Crumbs complex is largely unknown. The Crumbs complex is highly conserved between Drosophila and mammals (for reviews, see Medina et al., 2002; Roh and Margolis, 2003), and we and others have shown that Crb3 is involved in the building of tight junctions and in the control of apical morphogenesis in MDCK cells, a renal epithelial cell line (Lemmers et al., 2004; Roh et al., 2003). Pals1 (the mammalian homologue of Stardust) is also important for the establishment of tight junctions (Roh et al., 2002b) but its knockdown also induced a marked downregulation of PATJ making it difficult to attribute the effects observed to Pals1 knockdown alone (Straight et al., 2004).

To understand the role of PATJ in the Crumbs complex we used RNAi technology to produce PATJ knockdown (KD) stable epithelial cells derived from Caco2, a human intestinal polarized cell line. We show that strongly reduced expression of PATJ resulted in mislocalization of two tight junction proteins, occludin and ZO-3, to the lateral membrane without a loss of epithelial polarity. This could be explained by the fact that other tight junction-associated proteins, such as ZO-1 are not mislocalized. In PATJ KD cells, Pals1 is also no longer associated with tight junctions whereas Crb3, normally found at the apical membrane, in tight junctions and in small intracellular vesicles, is found in a large vesicular compartment that overlaps with early endosomes. These data strongly indicate that PATJ provides a link between lateral (occludin and ZO-3) and apical (Pals1 and Crb3) components of tight junctions and acts as a stabilizer of the Crb3 complex.

A clone of Caco2 (TC7 cells) was grown as described (Chantret et al., 1994). Cells were transfected using the Amaxa device, T solution and T20 program (Amaxa Biosystems, Germany) by mixing 3-5 μg plasmid with 100 μl buffer T and 1.5×10⁶ freshly trypsinized cells. Cells were selected with 2 mg/ml G418 for 1 week, 1 mg/ml for another week and then 0.5 mg/ml until single colonies could be picked with metal cloning rings. Colonies were amplified and tested for loss of expression of PATJ by immunofluorescence and western blotting. Cells (300,000) were seeded on Transwell filters (12 mm diameter, Corning, NY) and TER was measured every day with a MilliCell apparatus (Millipore Corporation, Bedford, MA).

The siRNA against PATJ were designed and produced by in vitro transcription with the Silencer siRNA kit from Ambion (Austin, TX). The DNA (5′-AACCAGATACTCACACTTCAG-3′) corresponding to one of them (siRNA-2) was introduced as a hairpin behind a U6 promoter added to peGFP-N2 (BD Bioscience Clontech, Palo Alto, CA) containing a neomycin-resistance cassette, using the HindIII and BglII restriction sites (a kind gift from M. J. Santoni, INSERM U599, Marseilles, France) and sequenced.

Antibodies against PATJ and Crb3 have been described (Lemmers et al., 2002; Lemmers et al., 2004). Chicken polyclonal antibodies SN47II against Pals1 were produced by J. Wijnholds (NORI, Amsterdam, The Netherlands) and rabbit polyclonal antibodies against ZO-1, 2 and 3 were kindly provided by K. Matter (University College London, UK). Rabbit polyclonal antibodies against aPKC and claudin1 were from Santa Cruz Biotechnology (Santa Cruz, CA) and Zymed laboratories (South San Francisco, CA), respectively. A. Quaroni (Cornell University, Ithaca, NY) provided mouse monoclonal antibodies against DPPIV and SI, and I. Trowbridge (Salk Institute, CA) provided mouse monoclonal antibodies against TFr. H.P. Hauri (Biozentrum, Basel, Switzerland) provided mouse monoclonal antibodies against Giantin. Mouse monoclonal antibodies against ESA (Ag525) and Lamp2 (Ac17) have been described (Le Bivic et al., 1988; Nabi et al., 1991). Mouse monoclonal antibodies against E-cadherin, EEA1, rab11 and occludin were purchased from BD Bioscience Transduction Laboratories (Palo Alto, CA) and Zymed (CA), respectively. Mouse monoclonal antibody against α-tubulin was from Sigma (St Louis, MO).

For immunofluorescence experiments, clones of Caco2 cells were seeded at near confluency on Transwell filters (24 mm in diameter, Corning, NY, USA), maintained for 8-10 days in culture and processed for immunofluorescence as described (Roh et al., 2003). Images were taken using a Zeiss 510 Meta confocal microscope (Zeiss, Le Pecq, France). Ultra-thin sections were prepared as published (Brandstatter et al., 1999) using confluent cells. Briefly, specimens were rapidly frozen in ethane and then transferred to a cryosubstitution unit EMAFS (LEICA Systems, Germany). Sections were incubated overnight at 4°C with the primary antibodies in 2% normal goat serum in 0.05 M Tris-buffered solution, pH 7.6. After washing, the sections were incubated with goat anti-rabbit or anti-mouse antibodies coupled to 6 nm or 15 nm gold particles, respectively in the same buffer. Samples for electron microscopy were prepared and processed as described (Breuza et al., 2002). Sections were observed on a Zeiss 912 electron microscope (Zeiss, Le Pecq, France).

Caco2 cell extracts were prepared and analyzed by western blotting as described (Lemmers et al., 2002) with primary antibodies and the corresponding secondary antibodies coupled to peroxidase (Jackson laboratories, Baltimore, PA). Bands were revealed using the Lumi light kit (Roche Diagnostics, Meylan, France) and quantified using the MacBas 2.5 (FujiFilm) software.

To decrease endogenous levels of PATJ in Caco2 cells, we have designed six siRNAs from the sequence of a human cDNA (Lemmers et al., 2002). These siRNAs were transiently transfected into Caco2 cells and their effect on PATJ expression was assayed by western blotting on Caco2 lysates. Using an affinity-purified rabbit antibody (Lemmers et al., 2002), one siRNA showed a marked decrease of the western blot signal. Stable clones were derived from Caco2 cells transfected with a plasmid containing neomycin resistance and a shRNA that reproduced the siRNA2 sequence. Knockdown clones for PATJ were screened by immunofluorescence and three clones (Cl4, 5 and 12) were selected together with a control clone (Cl8) containing the empty plasmid. PATJ was detected as several doublets (Lemmers et al., 2002) and upon siRNA2 expression there was a strong reduction in clones 4 and 12 (more than 90%) of the doublets migrating at 250, 190 and 140 kDa (Fig. 1A). Two bands were still detected: the 130 and 80 kDa bands (Fig. 1A and not shown for the 80 kDa band). The 130 kDa band was not found in PATJ immunoprecipitates indicating that it might be a protein with a common epitope only recognized after SDS denaturation whereas the 80 kDa band could either represent a crossreacting product or a protein produced by different transcripts not targeted by siRNA2. Immunofluorescence also confirmed the strong diminution of PATJ expression in clones 4 and 12 with the resultant loss of the typical junction pattern observed in the control clone 8 (Fig. 1B). No positive cells were observed in clone 4 and very few in clone 12 (<10%). In contrast, clone 5 exhibited only a slight reduction of PATJ signal on western blots (see Fig. 1A) and only patches of cells were downregulated (not shown) leading to the conclusion that the expression of the siRNA was not efficient in this clone. These combined observations strongly suggest that there was a significant (more than 90%) knockdown of PATJ in clones 4 and 12. Therefore we used these two clones to study the role of PATJ in epithelial cells.

We first investigated the effect of PATJ knockdown on the expression of Pals1 as it has been described that knockdown of Pals1 by siRNA in MDCK cells induced a strong inhibition of PATJ expression (Straight et al., 2004). Pals1 was detected and semiquantified after immunoblotting of cell lysates with antibodies against Pals1 (Fig. 2A). No significant difference (after normalization to the α-tubulin signal) was observed between control cells and PATJ KD cells indicating that PATJ was not crucial for maintaining normal cellular levels of Pals1 in Caco2 cells. We next asked whether the strong reduction of PATJ endogenous levels had any consequence on Pals1 organization. Control (Cl8) or PATJ KD (Cl4 and 12) Caco2 cells were grown on Transwell filters for 8 to 10 days and double indirect immunofluorescence was performed with antibodies against Pals1 and PATJ (Fig. 2B and C). In the control or the few PATJ-positive cells found in Cl12, Pals1 colocalized with PATJ at the level of tight junctions. In contrast, Pals1 was no longer accumulated on the apical membrane and at tight junctions of PATJ KD cells establishing that PATJ is indeed required for Pals1 localization and stabilization to tight junctions. In PATJ KD cells, Pals1 was however still associated to membranes (both plasma membrane and intracellular). Conversely, we transiently overexpressed a Myc-tagged mouse Pals1 (Roh et al., 2002b) in Caco2 cells prior to confluence to test its influence on PATJ localization. After 3 days of confluence, PATJ and Myc-Pals1 were localized by indirect immunofluorescence. PATJ was still concentrated at tight junctions but the overexpressed Myc-Pals1 was found all over the apical membrane, indicating that Pals1 alone is not sufficient to relocalize PATJ (Fig. 7A).

We next evaluated the impact of PATJ knockdown on polarity and junction markers. Confluent PATJ KD cells showed neither mislocalization of apical markers such as sucrase-isomaltase (SI), dipeptidylpeptidase IV (DPPIV), alkaline phosphatase (PLAP) or aPKC, nor of basolateral markers such as epithelial specific antigen (ESA) or E-cadherin (Fig. 3A for aPKC, DPPIV and E-cadherin, Fig. 5 for SI and not shown for the others) indicating that in PATJ KD cells, membrane polarity was not severely affected under these culture conditions. Several tight junction markers were also used and we did not observe a change in the distribution of ZO-1 or claudin 1 in PATJ KD cells (Fig. 3A for ZO-1). We investigated in parallel the functionality of tight junctions in PATJ KD cells compared to control cells by measuring the transepithelial resistance (TER) of monolayers during the first 4 days of culture on filters. No difference was observed, indicating that the barrier function of tight junctions was not altered 24 hours after cell-cell contacts and beyond (Fig. 3B). These data were supported by the fact that we did not see a change in the organization of tight junctions in PATJ KD cells by classical electron microscopy (Fig. 3C).

However, occludin and ZO-3 did show a perturbed pattern of localization in PATJ KD clones. In confocal Z sections, instead of being accumulated just below the spot of PATJ as in control cells, occludin and ZO-3 were localized along the lateral membrane (Fig. 4A,B, arrows). This aberrant accumulation of occludin or ZO-3 along the lateral membrane was also detected in X-Y confocal sections most often at the site of contact between several cells (Fig. 4B, arrow). We confirmed this delocalization of occludin and ZO-3 by immunogold labeling on ultra-thin sections (Fig. 4C). ZO-3 was occasionally detected in desmosome-like structures, an obvious sign of ectopic localization. Thus PATJ accumulation is necessary to restrict occludin and ZO-3 to tight junctions.

We next examined the effect of loss of PATJ expression on Crb3 subcellular distribution. In control cells, Crb3 is enriched in tight junctions and in the apical membrane, and is also found in numerous small intracellular vesicles (Fig. 5). Colocalization experiments using double-indirect immunofluorescence with known markers of Golgi (Giantin), basolateral recycling endosomes (TfR), late endosomes and lysosomes (LAMP2), apical recycling endosomes (rab11) and early endosomes (EEA-1) only showed occasional colocalization with rab11 (data not shown) and more consistently with EEA-1, leaving the localization of the majority of Crb3 intracellular labeling unknown (see Fig. 6B for EEA-1, arrows).

Strikingly, in PATJ KD cells, there was less Crb3 staining in the small cytoplasmic vesicles, instead Crb3 strongly accumulated in vesicular structures and overlapped with EEA-1, which in Caco2 cells is mainly found in vesicular structures just below the apical membrane (Fig. 6A,B, arrowheads and arrows). This colocalization of Crb3 and EEA-1 was confirmed by immunogold labeling on ultra-thin sections (Fig. 6C). Thus loss of PATJ expression induced a redistribution of Crb3 in the early endosomal network indicating that one of the functions of PATJ could be to prevent Crb3 accumulation in this compartment. This change in Crb3 distribution had no consequence on its expression level, which we confirmed by western blotting (not shown). To investigate further the relationship between Crb3 and PATJ and their respective subcellular localization, we overexpressed a VSV-G-tagged version of hCrb3 (VSV-G-Crb3) in Caco2 cells and observed a marked delocalization of aPKC but not of PATJ (Fig. 7B and C) indicating that, similar to Pals1, Crb3 alone is not able to relocalize PATJ. Our results imply that PATJ relies on other membrane cues for its proper localization.

To date, the Crb3 complex is known to consist of three partners, the Crb3 transmembrane protein, the MAGUK Pals1 and the poly PDZ-containing protein, PATJ. In Drosophila both Crb and Sdt, the respective homologues of Crb3 and Pals1 are essential for the maintenance of adherens junctions and thus epithelial polarity during gastrulation (for a review, see Knust et al., 1993). The fact that crb and sdt genes are expressed zygotically and not maternally facilitates the analysis of these mutant phenotypes. DPATJ (the homologue of PATJ) is already accumulated at the onset of cellularization (Bhat et al., 1999) and a recent study has shown that the multi-PDZ-containing protein homologous to PATJ is not involved in the dlt phenotype (Pielage et al., 2003) as previously suggested. These authors showed that the complementation of the deleted region that encompasses four genes (including dlt and Dpatj) rescued the dlt phenotype even in the absence of Dpatj. They therefore concluded that the Dpatj gene is not essential for epithelial polarity and development. As there are no known mutants for Dpatj, the exact role of this gene during epithelial polarity establishment and maintenance has not been established.

In mammals, the role of the Crb complex seems to be conserved: mutations in Crb1 result in a range of retinal dystrophies (den Hollander et al., 1999; Mehalow et al., 2003; van de Pavert et al., 2004) and overexpression of Crb3 or PATJ induces defects in both cell polarity and junction formation (Lemmers et al., 2004; Roh et al., 2003). However in both mammals and Drosophila, several questions about the cellular localization and function of the Crb complex remain unanswered. One of these is the precise cellular role of PATJ. In Drosophila, DPATJ is already localized at the level of basal adherens junctions before the Crb and Sdt genes are expressed (Knust and Bossinger, 2002; Tepass et al., 2001). Therefore it may interact with an, as yet, unidentified protein located at these junctions. In order to improve our understanding of the role of PATJ in establishing and maintaining tight junctions and cell polarity, we have used RNA interference in a human intestinal cell line Caco2.

Our results clearly show that PATJ is required for the proper localization of Pals1 to tight junctions. This was unexpected as Pals1 binds directly to Crb3, and this interaction could be thought sufficient for its localization. The requirement of PATJ to localize Pals1 may be explained by the fact that we also found that PATJ expression is necessary to restrict occludin and ZO-3 to tight junctions. Occludin binds directly to ZO-1, -2 or -3 (reviewed by Gonzalez-Mariscal et al., 2003; Matter and Balda, 1999) whereas PATJ binds to ZO-3 and claudin-1 (Roh et al., 2002a). Intriguingly, deletion of PATJ PDZ domain 6 that is involved in the interaction with ZO-3 prevented localization of an exogenous epitope-tagged PATJ (Roh et al., 2002a) whereas overexpression of full-length PATJ induced a loss of ZO-3 accumulation at tight junctions (Lemmers et al., 2002). Thus PATJ could serve as a stabilizer of occludin through its interaction with ZO-3 in tight junctions and both PATJ and ZO-3 might depend on each other to be properly localized. Similarly, PATJ may be required to stabilize the Crb3 complex to tight junctions. As neither occludin nor ZO-3 is identified in Drosophila, it will be interesting to determine whether the early expression of the Dpatj gene in the fly embryo is necessary for the correct localization of Sdt.

Localization of occludin to tight junctions is also regulated by phosphorylation (Wong, 1997) either by Src (Kale et al., 2003) or protein kinase C (Andreeva et al., 2001) but we have no evidence so far that PATJ regulates occludin phosphorylation and that it could act through this mechanism. It has been shown in a previous study that overexpression of a dominant-negative PATJ not only induced mislocalization of Pals1 but also of the Par3/6/aPKC complex and ZO-1 (Hurd et al., 2003) indicating that PATJ could play a general role in the building of tight junctions. This effect of the overexpression of a dominant negative was much stronger than that observed in PATJ KD cells, as we did not detect a change in aPKC or ZO-1 localization (Fig. 3A and not shown). These differences in the extent of perturbation of junction or polarity complexes might be attributed to a residual contribution of PATJ in our KD cells and/or to a non-specific effect of an overexpressed dominant-negative mutant of PATJ, which could bind to new (and non physiological) partners by its PDZ domains.

In contrast to occludin and ZO-3, there was no alteration in the localization of other tight junction proteins such as Claudin 1 and ZO-1 in PATJ KD cells. The organization of tight junctions in control and PATJ KD cells by electron microscopy appeared similar (see Fig. 3C), as was the transepithelial resistance during monolayer maturation on Transwell filters (see Fig. 3B). Together these data indicate that the overall structure of tight junctions was preserved in PATJ KD cells. As it is known that ZO-1 binds to ZO-3, it is at first puzzling that ZO-3 but not ZO-1 is delocalized in PATJ KD cells. This however could be explained by the fact that ZO-1 can associate either to ZO-2 or ZO-3 (Gonzalez-Mariscal et al., 2000) and we observed no delocalization of ZO-2 (not shown), indicating that in Caco2 cells most ZO-1 is in complex with ZO-2 instead of ZO-3. This however needs to be investigated in the future.

It therefore seems that in mammals, there is a sufficient redundancy to protect tight junctions from the loss of some of their constituents. This fact has been exemplified by several published studies. For example, in the occludin knockout mouse, there was no change in the barrier function (Saitou et al., 1998) and loss of ZO-1 expression does not prevent normal junction function or morphology or even epithelial polarity (Umeda et al., 2004). Similarly, downregulation of cingulin, another protein associated with tight junctions, has no effect on tight junction morphology or function but does appear to regulate the expression of several other genes (Guillemot et al., 2004).

PATJ expression also regulates Crb3 distribution in intracellular compartments such as the early endosomes. Crb3 redistribution was a direct consequence of PATJ downregulation and not a general effect on apical endocytosis since two other apical membrane markers (SI and DPPIV) were not accumulated in the EEA-1-positive compartment (Fig. 3A for DPPIV and Fig. 5 for SI). EEA-1 is found both in apical and basolateral early endosomes (Tuma et al., 2001) but given the apical localization of Crb3 and its sub-apical accumulation in PATJ KD cells, it is likely that it accumulates in apical early endosomes. Early apical endosomes are connected to the sub-apical compartment (SAC) that is known to be an intersection for many trafficking pathways in epithelial cells and a major sorting platform (reviewed by van der Wouden et al., 2003). We could not however measure Crb3 endocytosis as no antibodies are available against the short and highly O-glycosylated extracellular domain of Crb3 and Crb3 cannot be metabolically labeled because of its lack of cysteine or methionine residues. In addition, expression of a Crb3, tagged in the extracellular domain, induced defects in tight junction formation, apical membrane morphology (Lemmers et al., 2004) and epithelial polarity (Roh et al., 2003). Thus, overexpressed Crb3 cannot be used to study its normal trafficking in epithelial cells (see also Fig. 7B for the effect on aPKC, for example). A change in Crb3 subcellular distribution could have led to defects in apical membrane biogenesis but we did not observe such defects. It is thus likely that in PATJ-KD cells, there is still enough Crb3 present on the apical membrane and at tight junctions to prevent such defects. The precise role of Crb3 in apical membrane biogenesis has still to be investigated, but Fan and colleagues who have studied the role of Crb3 in MDCK cells using a knockdown approach did not report any polarity defect besides cilia morphogenesis (Fan et al., 2004). Furthermore, in Crb3 knockdown MDCK cells, PATJ is still correctly localized to tight junctions reinforcing the hypothesis that PATJ localization might rely on other partners.

Thus our results demonstrate that PATJ is required to stabilize both Crb3/Pals1 on the apical side of tight junctions and occludin/ZO-3 on the basal side of tight junctions. PATJ itself is was detected in the upper part of tight junctions by both immunofluorescence and immunoelectron microscopy (Lemmers et al., 2002) supporting the hypothesis that it might act as a determinant of the border between apical and lateral components. Loss of PATJ expression reveals that tight junctions in mammals are heterogeneous structures made of apical (the Crb3 complex and the Par6/aPKC complex) and lateral proteins (occludin, claudins, ZOs) and that PATJ acts as a scaffold to localize and stabilize these proteins. The fact that these insights into the role of PATJ were obtained using RNA interference in mammalian cells rather than from Drosophila genetics, highlights the power of this technology to dissect the role of each member of polarity complexes in cultured epithelial cells.

After this work was submitted, a paper was published which describes the effects of PATJ-KD in MDCK cells (Shin et al., 2005). In these cells, reducing the amount of PATJ led to defects in tight junction biogenesis and functionality and in epithelial organization when cultured in collagen gels. Although this is in good agreement with our hypothesis that PATJ is involved in tight junction formation and regulation, the phenotypic differences observed in our study might point at tissue-specific defects as MDCKII cells are derived from kidney and Caco2 cells from colon. Further work will clarify whether PATJ displays tissue-specific requirements for its cellular role.

We would like to thank J. P. Chauvin (IBDM, Marseilles) for his help with the immuno-electron microscopy and P. Rashbass (University of Sheffield, UK) for critical reading of the manuscript. We thank all the scientists who provided antibodies or plasmids and the technical platforms of the IBDM. This work was supported by grants from the European Community (QLG3-CT-2002-01266), Fondation de France (to A.L.B.), Cancéropôle PACA (to A.L.B.), ARC 4776 (to A.L.B.), ACI BCMS (French ministry for science) (to A.L.B.), Ligue Nationale contre le Cancer (to A.L.B.) and CNRS 6156. D.M. was supported by QLG3-CT-2002-01266.