EFA6 Facilitates the Assembly of the Tight Junction by Coordinating an Arf6-dependent and -independent Pathway*

We have previously reported that EFA6, exchange factor for Arf6, is implicated upon E-cadherin engagement in the process of epithelial cell polarization. We had found that EFA6 acts through stabilization of the apical actin ring onto which the tight junction is anchored. Mutagenesis experiments showed that both the catalytic domain of EFA6 and its C-terminal domain were required for full EFA6 function. Here we address the contribution of the specific substrate of EFA6, the small G protein Arf6. Unexpectedly, depletion of Arf6 by RNA interference or expression of the constitutively active fast-cycling mutant (Arf6T157N) revealed that Arf6 plays an opposing role to EFA6 by destabilizing the apical actin cytoskeleton and the associated tight junction. However, in complementation experiments, when the C-terminal domain of EFA6 is co-expressed with Arf6T157N, it reverts the effects of Arf6T157N expressed alone to faithfully mimic the phenotypes induced by EFA6. In addition, we find that the two signaling pathways downstream of EFA6, i.e. the one originating from the activated Arf6GTP and the other one from the EFA6 C-terminal domain, need to be tightly balanced to promote the proper reorganization of the actin cytoskeleton. Altogether, our results indicate that to regulate the tight junction, EFA6 activates Arf6 through its Sec7 catalytic domain as it modulates this activity through its C-terminal domain.

We have previously reported that EFA6, exchange factor for Arf6, is implicated upon E-cadherin engagement in the process of epithelial cell polarization. We had found that EFA6 acts through stabilization of the apical actin ring onto which the tight junction is anchored. Mutagenesis experiments showed that both the catalytic domain of EFA6 and its C-terminal domain were required for full EFA6 function. Here we address the contribution of the specific substrate of EFA6, the small G protein Arf6. Unexpectedly, depletion of Arf6 by RNA interference or expression of the constitutively active fast-cycling mutant (Arf6T157N) revealed that Arf6 plays an opposing role to EFA6 by destabilizing the apical actin cytoskeleton and the associated tight junction. However, in complementation experiments, when the C-terminal domain of EFA6 is co-expressed with Arf6T157N, it reverts the effects of Arf6T157N expressed alone to faithfully mimic the phenotypes induced by EFA6. In addition, we find that the two signaling pathways downstream of EFA6, i.e. the one originating from the activated Arf6GTP and the other one from the EFA6 C-terminal domain, need to be tightly balanced to promote the proper reorganization of the actin cytoskeleton. Altogether, our results indicate that to regulate the tight junction, EFA6 activates Arf6 through its Sec7 catalytic domain as it modulates this activity through its C-terminal domain.
The internal bodily cavities are covered by a monolayer of epithelial cells that provides the first line of protection to pathogenic agents by acting as a physical barrier and by supporting mucosal immunity. Epithelial tissues are submitted to major remodeling during which epithelial cells are turned into mesenchymal cells along a process named epithelial-mesenchymal transition. Conversely, mesenchymal cells differentiate into epithelial cells in a process named mesenchymal-epithelial transition. Both processes take place during normal development and pathologic conditions. These phenotypical transition events are not supported by one unique process but rather refer as to a large variety of cell plasticity phenomenon in response to various pathways. One major challenge in modern cell biology, especially oncology, is to decipher the molecular mechanisms underlying these biological events (1)(2)(3).
Polarized epithelial cells present two distinct plasma membrane domains; they are the apical domain exposed to the lumen and a basolateral domain contacting the extracellular matrix and the neighboring cells. These two domains are separated by the most apical junctional complex named the tight junction (TJ) 3 that is a hallmark of epithelial cell polarity. The TJ serves several functions; a barrier function to control the paracellular transport, a fence function to prevent the intermixing of proteins and lipids of the outer leaflet of the apical and basolateral domains, delivery site for exocytotic vesicles, and finally a signaling platform for the regulation of cell polarity, proliferation, and differentiation (4). The tetraspan proteins of the claudin family are the basic molecular units of the TJ. Their two extracellular loops capable of transcellular homo-and heterodimerization link the TJ of contacting cells to form the physical junction. Occludin and junctional-associated molecule are two other transmembrane proteins present in the TJ which roles are still elusive (5). Another hallmark of polarized epithelial cells is the organization of the actin cytoskeleton that comprises several distinct structures including an apical thick ring of actomyosin that supports and accommodates the TJ. The cytosolic proteins of the ZO family (ZO-1, ZO-2, and ZO-3) bind to the cytoplasmic tails of the TJ integral membrane proteins and serve as a link to associate the TJ to the underlying apical actin ring (6).
Much needs to be done to uncover the interconnecting signaling pathways underlying epithelial-mesenchymal transition and mesenchymal-epithelial transition (1,7). The small G protein Arf6 was shown to affect the turnover of the adherens junction, cell-cell adhesion, cell migration, tumor cell invasion, and hepatocyte growth factor-induced tubulogenesis in the epithelial cell culture model Madin-Darby canine kidney (MDCK). As such, Arf6 has been proposed to function as a critical determinant of the epithelial to mesenchymal transition (8 -14). Several GAPs and GEFs have been described to control Arf6 GTP/ GDP cycle as well as the morphology of epithelial cells. Of interest, some ArfGAPs have been found to function not only as GTPase activating protein but also as Arf GTP-bound protein effectors (15,16). Among the GEFs identified for Arf6, we have characterized EFA6 as a potent regulator of TJ remodeling in MDCK cells (17). Other ArfGEFs such as ARNO and BRAG2/ GEP100 have been proposed to function through Arf6 and to affect the cortical actin cytoskeleton (18 -20). Because any given cell type expresses several Arf6GEFS and Arf6GAPs, an important question is how these Arf6 regulators are selectively enrolled and how they determine the biological effects of Arf6.
Our mutagenesis analysis of EFA6 indicated that its C-terminal domain, otherwise shown to rearrange the actin cytoskeleton in fibroblastic cell lines (21), and Sec7 domain which contains the catalytic nucleotide exchange activity were both essential for EFA6 to contribute to the TJ assembly (17). Here, we studied the contribution of Arf6 and found that the activated Arf6GTP acts in conjunction with the C-terminal domain of EFA6. To evaluate this cooperative mechanism, we conducted complementation experiments and showed that co-expression of a constitutively activated mutant of Arf6 and the C terminus of EFA6 phenocopied the effects of the full-length EFA6 in all our assays examining the actin cytoskeleton in nonpolarized and polarized cells and the formation of the TJ in polarized cells.

EXPERIMENTAL PROCEDURES
Cells and DNA-Baby hamster kidney (BHK) cells were grown in GMEM (Invitrogen) and MDCKII cells in MEM (Sigma-Aldrich) supplemented with 5% heat decomplemented fetal calf serum (Biowest-Abcys, Paris, France) and penicillinstreptomycin (Invitrogen). Transient transfections in BHK cells were performed using Dreamfect according to the manufacturer's instructions (OZBiosciences, Marseilles, France). MDCK-T23 cells stably expressing the tetracycline-controlled transactivator have been described elsewhere (22). The cDNAs encoding for the C-terminal HA-tagged Arf6WT, Arf6T27N, and Arf6T157N were introduced into the vector pUHD10-3 and expressed under the control of the tetracycline responding element. The shRNA-specific for canine Arf6 (target sequence gcaccgcattatcaatgaccg) was cloned into the pTER vector. The construct comprised of the PH and C terminus of ARNO triglycine (ARNO PHCter, residues 262-400) was cloned into pEGFP-C1, and the PHC-terminal domain of EFA6A was cloned into pEGFP-C3 (EFA6A PHCter, residues 349 -645) (23). All transfections in MDCK cells were performed using Lipofectamine 2000 (Invitrogen). We have only used clonal cell lines; several clones for each of the constructions have been studied and verified to give similar results. Cells were passaged in 10-cm dishes and induced to polarize on 12-mm, 0.4-m pore size Transwell polycarbonate filters (Corning-Costar, Cambridge, MA) in the presence or absence of 20 ng/ml doxycycline (ϩ/ϪDox). In all experiments, the expression of the Arf6 proteins or the shRNA was allowed by growing the cells in the absence of Dox for 36 -48 h before analysis.
Paracellular Transport Assay-The formation of the TJ barrier was assessed by first plating the cells on filters in normal MEM. The next day the cells were submitted to a calcium switch procedure described in detail elsewhere (17,25). Briefly, the cells were washed 3 times quickly in PBS and 3 times for 10 min under agitation in 2 mM PBS-EGTA. The cells were then incubated for 6 h in spinner medium (Sigma-Aldrich) supplemented with 5 M calcium (ϩ/ϪDox) to completely breakdown the TJ as controlled by measuring the transepithelial resistance. At t ϭ 0, the calcium switch was performed by replacing the medium with normal MEM (ϩ/ϪDox). At the indicated times (30, 60, and 90 min) TRITC-dextran 9kD (1 mg/ml in normal MEM) was added in the apical chamber (200 l), and 800 l of normal medium was added in the bottom chamber. 2 h later the total amount of TRITC-dextran in the basal medium was quantitated using a spectrofluorimeter. The graphs report the percentage recovered in the basal medium of the total amount of reagent added apically. The disassembly of the TJ barrier was assessed on fully polarized cells grown on filters for 5 days. The cells were washed 3 times quickly in PBS and three times for 10 min under agitation in PBS-EGTA 2 mM and further incubated in spinner medium supplemented with 5 M calcium (ϩ/ϪDox). At the indicated times (30, 60, 120, and 240 min) TRITC-dextran was added apically, and its paracellular transport was quantified as described above. All experiments were repeated at least three times using triplicates.
Transepithelial Electrical Resistance (TER) Measurement-The TER was measured as described previously using duplicates or triplicates for each measure (25). Results are expressed in ohms/cm 2 after subtraction of the TER obtained from a duplicate of empty filters.
Confocal Immunofluorescence-Cells were fixed by 4% paraformaldehyde on ice for 30 min rapidly rinsed in PBS, 1 mM calcium, 0.5 mM magnesium. The samples were then prepared as previously described (17). Images were processed for presentation using the NIH Image and Adobe Photoshop CS2 software. Images from the calcium switch experiments and latrunculin treatment are representative of at least three independent experiments.
Immunoblot-Samples were resolved on SDS-PAGE, and proteins were transferred onto a nitrocellulose membrane.
Membrane blocking and antibody dilutions were done in PBS, 5% nonfat dry milk. The proteins were revealed by chemiluminescence (ECL TM , Amersham Biosciences) using secondary antibodies directly coupled to horseradish peroxidase (Jackson ImmunoResearch Laboratories, Inc.). The membranes were analyzed with the luminescent image analyzer LAS-3000 (Fujifilm).
Biotin Endocytosis Assays-This experiment has been extensively described elsewhere (25). Briefly, cells grown on 12-mm filters and washed in ice-cold PBS, 1 mM calcium, 0.5 mM magnesium (PBS-CM) were cell surface-biotinylated in PBS-CM containing 1 mg/ml NHS-SS-biotin. The free biotin was neutralized, and the cells were incubated for different periods of time at 37°C in MEM-bovine serum albumin to allow for endocytosis. Except for control samples, the cells were then treated with glutathione (50 mM glutathione, 50 mM Tris, 100 mM NaCl, 0.2% bovine serum albumin, pH 8.6) to remove the remaining biotin from the cell surface. The cells were then solubilized in Triton-X100 lysis buffer, and the biotinylated proteins were recovered on streptavidin-agarose beads. The eluted proteins were resolved on SDS-PAGE and revealed by immunoblot. All experiments were repeated at least three times using triplicate or quadruplicate samples. Quantitation was performed after scanning of non-saturating signals and using different time exposures with the luminescent image analyzer LAS-3000 (Fujifilm).

Arf6 Impacts the Assembly and the Stability of the Tight
Junction-To address the role of Arf6 on epithelial cell polarity, we analyzed the effects of its depletion on the TJ and the actin cytoskeleton. By transient transfection we have tested several small interfering RNAs specific for the canine mRNA sequence of Arf6. Control small interfering RNAs (siRNAs) including a scramble sequence and two siRNAs specific for the human Arf6, each with one mismatch to the canine sequence (supplemental data), had no effect on endogenous Arf6 expression. Using a stable cell line in which an Arf6-specific shRNA was expressed under the control of the tetracycline transactivator, we found that 48 h after induction of the expression of the shRNA (ϪDox) the levels of the endogenous Arf6 were markedly reduced. The expression levels of the related proteins Arf1 and Arf5, of the exchange factor EFA6, and the major constituents of the adherens and tight junctions were unaffected (supplemental data).
Then we examined the formation of the TJ during epithelial cell polarity development using a calcium switch procedure. The barrier function of the TJ was assessed by measuring the passive transport of fluorescent TRITC-dextran (9 kDa) from the apical to the basal compartment (see "Experimental Procedures"). The more TRITC-dextran accumulates in the basal medium, the slower the TJ barrier is established. When Arf6 expression was repressed (ϪDox) the gain of the permeability barrier was accelerated compared with control cells (ϩDox) as judged by lower levels of TRITC-dextran that diffused across the monolayer (Fig. 1A). Likewise, upon calcium depletion the loss of the permeability barrier of a fully polarized MDCK monolayer was delayed as indicated by the slower appearance of TRITC-dextran in the basal medium (Fig. 1B). In parallel, we measured the TER and found similar results ( Fig. 1, C and D). These functional observations were confirmed at the morphological level by confocal immunofluorescence looking at the TJ-associated cytosolic protein ZO-1 and the apical actin ring. Depletion of Arf6 accelerated the appearance of ZO-1 and filamentous actin to the newly forming TJ. There are several aspects of the cell that we observed; for example, the proportion of the plasma membrane stained for ZO-1 and its underlying cortical actin and the shape of the cell, where a more cuboidal shape is consistent with epithelial cell polarization. We found that depletion of Arf6 accelerated the appearance of ZO-1 and filamentous actin to the newly forming TJ. A larger part of the periphery of the shARf6 cells was stained for ZO-1 and phalloidin, and these cells maintained an epithelial cuboid shape. Similarly, the dissipation of the ZO-1 and phalloidin staining from mature TJ was delayed when calcium was withdrawn. These observations were unexpected as we had shown that EFA6 overexpression, which increases the levels of the active GTPbound Arf6, had the similar effect of accelerating the assembly of the TJ. Thus, we further studied the role of Arf6 by analyzing the effects of the exogenous expression of the wild-type Arf6 (Arf6WT), the dominant negative mutant Arf6T27N, and the constitutively activated fast-cycling mutant Arf6T157N. Note that the induced expression (ϪDox) of the exogenous Arf6 proteins was kept near or below the one of the endogenous Arf6 (see supplemental data and "Discussion"). Exogenous expression of Arf6WT affected neither the formation nor the disassembly of the TJ at the functional and morphological levels ( Fig.  1). Expression of the dominant negative Arf6T27N mutant delayed the formation of the TJ and accelerated their disassembly. This is coherent as Arf6T27N is believed to exert its dominant negative effect by sequestering its cognate exchange factor(s), notably EFA6 (26), and we found that EFA6 stabilizes the TJ (17). We cannot exclude that other Arf6GEFs such as ARNO and BRAG2/GEP100 might also be sequestered by Arf6T27N.
The constitutively activated mutant, Arf6T157N, exhibited a disruptive action on the TJ. Similar results were obtained with the GTP-locked mutant Arf6Q67L (data not shown). Although unexpected with respect to EFA6 stimulatory effects, this result was in agreement with our observation upon Arf6 depletion where TJ stabilization was increased. At steady state, in contrast to all the other cell lines, both Arf6T27N-and Arf6T157Nexpressing cells displayed a leaky TJ phenotype even after 7 days grown on filters in normal medium. This is reflected in the TER data (see Figs. 1 and 3).
To summarize, on one hand the Arf6T27N data supported the positive impact of EFA6 on the TJ, but on the other hand we found that Arf6 conveys negative effects (shRNA and Arf6T157N). To reconcile these apparently opposite results, we reconsidered the role of the C-terminal domain of EFA6 that we had previously found to be also required to accomplish EFA6 positive effects on the TJ. We postulated that it is the combined action of the EFA6-activated Arf6GTP and the C-terminal domain of EFA6 that are necessary to recapitulate faithfully the observed EFA6 effects on the TJ. Non-polarized Cells-To test this hypothesis we performed complementation experiments where the cells have been co-transfected with the constitutively activated mutant Arf6T157N and the C-terminal domain of EFA6. To ensure proper localization at the plasma membrane of the C-terminal domain, we used the PH C-terminal (PHCter) region of EFA6 that includes the adjacent EFA6 PH and C-terminal domains. First, we analyzed the effects on the actin cytoskeleton of BHK cells in which we had previously studied in detail the impact of the expression of the full-length and truncated EFA6 proteins (21,27). Cells transfected with EFA6 display on their dorsal surface numerous microvilli-like structures that represent short filopodia. At the periphery, few small membrane ruffles often unattached to the support were observed. EFA6 is found concentrated in both of these peripheral membrane structures where it co-localizes with filamentous actin ( Fig. 2A  There is no apparent alteration of the dorsal plasma membrane that presents only few microvilli-like structures. Filopodia are rarely observed at the periphery of the membrane that appears smooth. The quantification showed that 85% (28 cells of 33) of the cells displayed a classical Arf6T157N phenotype, 9% (3/33) that were lowexpressing were undistinguishable from non-transfected control cells, and 6% (2/33) resembled to EFA6 transfected cells (see the Fig. 2D  graph). Expression of the PHCter construct of EFA6 led to the formation of filopodia, which depending on the levels of expression, can be small and numerous at the periphery and/or appeared as long protrusions that deform the cell shape ( Fig.  2A). The PHCter is preferentially localized in these membrane extensions rich in polymerized actin. On the dorsal surface few filopodia always of longer length than in nor- E and F, confocal IF analysis at the TJ levels of the indicated cell monolayers after calcium switch (t ϭ 120 min) or calcium removal (t ϭ 120 min), respectively. Cells were co-stained for ZO-1 and the filamentous actin (phalloidin). mal cells can be observed. All of the transfected cells (28/28) had this distinctive phenotype. We also studied the exchange factor ARNO that displays a shorter C-terminal domain (25 residues versus 150 residues for EFA6). Expression of the PHCterminal domain of ARNO had no apparent effect on the morphology of the cells. The protein was found throughout the cytosol and was absent from the plasma membrane ( Fig. 2A) in agreement with previous observations (28).
When Arf6T157N and EFA6 PHCter were co-expressed, a majority of the cells resembled EFA6-expressing cells. As an example, in Fig. 2B the four left panels show a cell that presents a typical EFA6 phenotype with a substantial co-localization of both proteins together with the filamentous actin. An enlarged image of another cell is shown to the right to illustrate the presence of peripheral ruffles and short filopodia, both membrane shapes typical of the EFA6 phenotype and not found in PHCter or Arf6T157N expressing cells. Although quantitative protein amounts cannot be determined, it is striking to note that the ratio of the fluorescent signals of the individual proteins is constant in all the pixels where co-localization occurs. To further quantify the complementation effect, we co-transfected different quantities of both plasmids to maintain a constant expression of the PHC-terminal domain of EFA6 and decreasing amounts of Arf6T157N as verified by immunoblot (Fig. 2B). However, by confocal immunofluorescence, individual cells displayed a wide range of relative levels of expression of the two proteins Arf6T157N and PHCter; thus, we have defined three groups of cells to rationalize the quantification. In group I, the cells had an Arf6T157N staining intensity at least three times stronger than the one of the PHCter; in group II the cells had staining similar or not exceeding twice that of the other marker, and in group III the intensity of the staining for PHCter was at least 3 times that of the one for Arf6T157N. The results of the quantification are shown in Fig. 2D. In group I, 72% (35/49) of the cells displayed a phenotype similar to EFA6-expressing cells, 10% (5/49 cells expressing low levels of both proteins) had an unchanged phenotype, and 18% (9/49 cells expressing very high levels of Arf6T157N) had a phenotype resembling the one induced by Arf6T157N expression. In group II, 20% (8/41) of the cells had the EFA6 phenotype, 17% were unchanged, 42% had a mixed phenotype with filopodia and lamellipodia on different places of the plasma membrane, 7% were typical of PHCter-transfected cells, and 5% looked like Arf6T157N cells. In group III, the cells displayed mostly the PHCter phenotype (88%, 15/17), whereas the rest had a mixed phenotype (12%, 2/17). The quantitative analysis of the different cell populations indicated the existence of a balance between the signaling pathways associated either to the activated Arf6 or to the C-terminal domain of EFA6. This equilibrium needs to be finely tuned experimentally to mimic the EFA6 regulated phenotype. Also, it is interesting to note that not only do Arf6T157N and the PHCter colocalize in the EFA6-like actin structures, but they do so at a constant ratio of their levels of expression, suggesting a stoichiometric equilibrium.
When ARNO PHCter was co-expressed with Arf6T157N at various ratios, it had very little impact on the phenotype  1 and 2) or co-transfected at different ratios (samples 3-6). D, BHK cells transfected for the indicated proteins were analyzed to quantitate the proportion of cells displaying a canonical EFA6 phenotype. Within the different co-transfected cell populations, we defined three groups of cells I, II, and III exhibiting different relative staining intensities for the indicated proteins (see "Results" for more detail). The quantification reports the values of a representative experiment where all the transfections have been carried out simultaneously and the samples analyzed side-by-side by two independent assessors. The experiments were confirmed twice for ARNO and at least three times with the other plasmids. OCTOBER 31, 2008 • VOLUME 283 • NUMBER 44

JOURNAL OF BIOLOGICAL CHEMISTRY 30133
induced by Arf6T157N and never reproduced the EFA6 phenotype (Fig. 2, B-D). We conducted a similar quantitation as performed with the EFA6 PHCter. In group I, 100% of the cells had an Arf6T157N phenotype. In group II, 75% (15/20) of the cells had an Arf6T157N phenotype, and the rest of the cells that were expressing low levels of both proteins displayed a normal morphology. In group III, 40% of the cells resemble Arf6T157Nexpressing cells, and 60% had a normal phenotype. The decrease of the percentage of cells displaying an Arf6T157N phenotype with increasing amounts of ARNO PHCter suggests that ARNO competes with the effector(s) of Arf6T157N. In fact, it is worth noting that in double-transfected cells, a significant amount of ARNO PHCter was found co-localized at the plasma membrane with Arf6T157N. These results are in agreement with the recent observation that Arf6GTP is capable of recruiting ARNO as an effector at the plasma membrane (29). In any case, these results show that only the C-terminal domain of EFA6, but not the one of ARNO, cooperates with the activated Arf6 to remodel the actin cytoskeleton.
Altogether, our results suggest a model whereby EFA6 is associated through the activation of Arf6 and its C-terminal domain to two signaling pathways that then act cooperatively to fulfill their action on the organization of the actin cytoskeleton. Our observations also indicate that these signals would be both necessary and sufficient and that the remainder of the protein located N-terminal to the Sec7 domain would not participate. We propose that, individually, the two pathways lead to very different, if not antagonistic, phenotypes, providing an explanation to our initial apparent contradictory observations of the effects of EFA6 and Arf6 on cell polarity development.
In Polarized Epithelial Cells Arf6 and the PHC-terminal Domain of EFA6 Cooperate to Mimic the Effects of EFA6 on the Formation and Maintenance of the TJ-We next asked whether this cooperative mechanism was utilized by EFA6 to contribute to epithelial cell polarity. To this aim, we made an MDCK cell line expressing the PHC-terminal domain of EFA6 in a constitutive manner and Arf6T157N under the control of the inducible Tet-off system. The cell line was selected to express a reasonable amount of both proteins in a ratio that is comparable with the cells mimicking EFA6 effects in non-polarized cells (see supplemental data). The expression of PHCter alone (cells grown in the presence of doxycycline, ϩDox) had no measurable effect on the acquisition of the barrier function as measured by the paracellular transport of TRITC-dextran and the TER (Fig. 3, A and C). As shown in Fig. 1, Arf6T157N delayed the formation of the TJ barrier. However, the expression of PHCter reverted the negative impact of Arf6T157N, and instead, expressed together, the two proteins accelerated the establishment of the permeability barrier mimicking the effect of EFA6. The morphological analysis by confocal immunofluorescence confirmed the cooperative action of the two proteins to promote the assembly of the TJ (Fig. 3, A and C, left panels). We also analyzed the stability of the TJ that we had found to be increased in EFA6-expressing cells. We observed that the PHCter had no effect on its own but could revert the destabilizing effect of Arf6T157N leading to a more stable TJ, similar to the EFA6 action when expressed alone (Fig. 3, B and D, left  panels).Because the EFA6 PHCter had no effect on its own but when expressed together with Arf6T157N induces a phenotype that is different from the one induced by Arf6T157N alone, one can rule out the possibility that the PHCter acts by sequestration of Arf6T157N. In conclusion, our results suggest that EFA6 promotes the formation and stabilizes the TJ through the cooperative effects of the Arf6GTP-and PHCter-associated signaling pathways.
The Co-expression of the PHC-terminal Domain of EFA6 and Arf6T157N Mimics the Molecular Action of the Full-length EFA6 on the TJ-associated Actin Cytoskeleton-We had shown that EFA6 acts upon the TJ by affecting the reorganization of the actin cytoskeleton, specifically stabilizing the apical actin ring onto which the TJ is anchored, thus leading to an increased retention of the TJ proteins at the cell surface (17). To assess whether the combined expression of Arf6GTP and the PHCterminal domain of EFA6 faithfully recapitulate the activity of EFA6, we looked in more detail at the actin cytoskeleton structural organization and its association to the TJ. We examined three molecular parameters that are selectively affected by EFA6; 1) the accelerated reorganization of the apical actin cytoskeleton during epithelial cell polarity development and its stabilization in fully polarized cells, 2) resistance to latrunculin treatment of the apical actin ring in polarized cells, and 3) the retention of TJ proteins at the plasma membrane cell surface of polarized cells.
In a commensurate manner to the TJ assembly, Arf6T157N expression significantly retarded the reorganization of the apical actin cytoskeleton and the gain of the epithelial cuboidal morphology compared with EFA6. Similarly, upon calcium withdrawal the apical actin cytoskeleton was rapidly disassembled, and the cells adopted a round shape (Fig. 3, C and D, right panels, see also Fig. 1, C and D, right panels). In contrast, the co-expression of PHCter turned the cells into EFA6-like cells that rearrange rapidly the TJ-associated actin cytoskeleton and displayed an epithelial morphology. Likewise, in polarized cells the apical actin cytoskeleton was only slightly affected after 30 or 60 min of incubation in the absence of calcium (Fig. 3, C and D, right panels).
We have previously shown that the exogenous expression of EFA6 renders the apical actin ring that supports the TJ selectively resistant to dissolution by latrunculin (17). After 15 min of exposure to latrunculin (5 M), the basolateral filamentous actin was almost completely dissolved in both control and EFA6-expressing cells. In contrast, the apical actin ring was markedly disassembled only in control cells but found intact in EFA6-expressing cells especially where TJ are present in between adjacent cells (Fig. 4). There was no detectable effects in cells expressing Arf6WT (data not shown). In contrast, expression of Arf6T157N rendered the cells highly sensitive to latrunculin with an almost complete loss of the apical actin ring. In cells stably expressing the PHC-terminal domain, there was no apparent effect compared with control cells. When Arf6T157N expression was induced in the PHCter stable cell line, the cells became resistant to latrunculin to the same extent as the EFA6-expressing cells. Thus, the expression of the PHCter domain reverted the hypersensitive latrunculin phenotype produced by Arf6T157N.

EFA6 Dual Signaling Pathways Regulate Tight Junction
A consequence of EFA6-induced stabilization of the apical actin cytoskeleton is the retention at the plasma membrane of the actin-associated transmembrane proteins forming the TJ (17). Thus, we examined whether the combined expression of Arf6T157N and EFA6 PHCter could similarly increase the lifetime at the cell surface of claudin 2 (Cld2), the major constituent of the TJ in MDCKII cells. This was done by cell surface biotinylation followed by a chase and cell stripping at the indicated times to quantitate the amount of remaining Cld2 at the cell surface (see "Experimental Procedures"). Expression of Arf6T157N considerably reduced the time spent by the Cld2 at the cell surface (Fig. 5). In contrast, the PHC-terminal domain had little effect, but the two combined increased the lifespan of Cld2 at the cell surface as previously found upon EFA6 exogenous expression (17). The same results were obtained for occludin (data not shown). As a control, we examined the fate of the epidermal growth factor receptor and observed no difference regardless of the proteins expressed. Altogether, these results indicate that the co-expression of Arf6T157N and EFA6 PHCter recapitulates the effects of EFA6 on the apical actin cytoskeleton and the TJ in epithelial polarized cells.

DISCUSSION
Although alternative explanations exist, we would like to propose that the results presented here indicate that EFA6 proceeds by the coordination of signals associated to two of its structural domains to regulate the TJ; (i) the catalytic Sec7 domain responsible for the activation of Arf6 and (ii) the C terminus of EFA6 for which partners have yet to be identified. Our approach consisted of complementation experiments with the two separate EFA6-associated entities involved: activated Arf6 and EFA6 PHCter. By this means, we could phenocopy all the modifications induced by the full-length EFA6; the effects For each cell line, data were compared between ϩ/Ϫdoxycycline by using a paired Student's t test (*, p Ͻ 0.001). E and F, confocal IF analysis at the TJ levels of the indicated cell monolayers after a calcium switch (t ϭ 120min) or calcium removal (t ϭ 120 min), respectively. Cells were co-stained for ZO-1 and the filamentous actin (phalloidin). OCTOBER 31, 2008 • VOLUME 283 • NUMBER 44 on the actin cytoskeleton in non-polarized cells, stimulation of TJ assembly, the rearrangement of the supporting apical actin cytoskeleton during epithelial cell polarization, stabilization of the TJ in fully polarized cells, reduced TJ proteins endocytosis, and resistance to the actin destabilizing drug latrunculin. Our results indicate that both signals are necessary and sufficient to control the TJ and that the N-terminal domain of EFA6 is not required.

EFA6 Dual Signaling Pathways Regulate Tight Junction
We have expressed Arf6 and EFA6 under the control of the tetracycline-repressible expression system to achieve low levels of expression and to avoid toxic and/or nonspecific effects due to overexpression. As illustrated by their impact on the morphology and the TJ, we find that the dominant negative Arf6T27N and the fast-cycling Arf6T157N exert significant effects. Given that the endogenous Arf6 is mostly found in the cells in a GDP-bound inactive state (26,27), its presence does not compete with the mutants. Likewise, at low levels of expression and as previously shown (21,23), the EFA6 PHCter has a dramatic impact on the morphology of the cells even in the presence of the endogenous protein.
Even though the endogenous EFA6 is expressed at very low levels, we speculate that EFA6 is maintained in an inactive conformation with the PHCter domain closed. The PHCter construct would then represent an open active form of EFA6. This would also explain why the expression of Arf6T157N alone but in the presence of the endogenous EFA6 did not mimic the effects of the co-expression Arf6T157N and EFA6 PHCter. The other reason, discussed hereafter, would be the need for a stoichiometric balance between the signals delivered by Arf6 and EFA6 PHCter.
As stated above, the co-expression of Arf6T157N and EFA6-PHCter led to an effect opposite to the one obtained by the expression of Arf6T157N alone. Thus, rather than operating through sequestration, to passively neutralize the detrimental effects of Arf6T157N on the TJ, we propose that EFA6-PHCter in the presence of Arf6T157N reproduces the full effects of EFA6. This observation supports the presence of a coordinated activity between the C-terminal domain of EFA6 and activated Arf6.
Our functional complementation experiments indicated that the strength of the signals delivered by the two proteins Arf6T157N and EFA6 PHCter needed to be finely equilibrated to mimic EFA6 effects. This stoichiometric control is surprising considering that Arf6 is catalytically activated and that no catalytic activity is known to be associated to the C-terminal domain. Nevertheless, when looking at the cell morphology or at the actin cytoskeleton, overexpression of EFA6 even at very high levels does not favor one pathway over the other, suggesting that EFA6 is at the focal point of the equilibrium.
A first explanation would be that the C terminus of EFA6 binds an enzymatic protein connected to a pathway that cooperates downstream from EFA6 with the one associated to Arf6GTP. The C terminus and Arf6GTP pathways lead to the formation of long membrane extensions or to large membrane ruffles, respectively. The coordination of both of these activities would facilitate cell-cell adhesion and TJ assembly as suggested by overexpression of EFA6 alone. We are currently searching for interactors of the C terminus of EFA6 to determine its mode of action. As to Arf6, its effects on the actin cytoskeleton are believed to be mediated by Rac (30), which has also been implicated in the development of MDCK polarity, TJ assembly, and organization of the apical actin ring (22,31,32). Arf6 might also play a role by modulating the phospholipid metabolism at the plasma membrane (30).
Another possibility, in agreement with our previous data (27), is that an Arf6GAP acts as the Arf6GTP effector, in which case it will neutralize de facto the amplification of Arf6 activation by EFA6. Indeed, there would be one converted Arf6GTP for one Arf6GAP that would cooperate with the associated EFA6 C terminus pathway. Besides, several Arf6GAPs have been proposed to contribute to Arf6GTP induced membrane ruffling (33,34).
Third, although not exclusively so, one can imagine that EFA6 serves as a platform in which the Arf6GTP effector associates directly or indirectly to the PHCter so that for each converted Arf6GTP there is a stoichiometric control by the C-terminal domain of EFA6. This model is further supported by the following observation; the co-localization of Arf6T157N and EFA6 PHCter is similar to the one we had observed for the Arf6 GDP-locked mutant Arf6T44N and the EFA6-PH domain in  discrete areas of the plasma membrane that resemble the dorsal short filopodia-like structures (26). One could speculate that EFA6, through its PH domain, and Arf6GDP are targeted independently to specialized areas of the plasma membrane where they interact and EFA6 catalyzes Arf6 nucleotide exchange. The converted Arf6GTP would then stay in the vicinity of EFA6 within a multimolecular complex that would contain Arf6GTP effector(s), including Arf6GAPs, and binding partners of the C terminus of EFA6. This functional complex would affect the actin cytoskeleton through a mechanism that remains to be determined.
In any case, our cooperative model implies that the role of Arf6 can only be understood in the context of its activation by EFA6 and helps to explain some confusing observations. First, although EFA6 is capable of remodeling the actin cytoskeleton, accelerate the TJ formation, and stimulate epithelial polarization, the activated Arf6 could not recapitulate any of these effects. On the contrary, its expression led to opposite effects with a slower formation of the TJ and higher instability, similar to the effects observed by others on the adherens junction (8). Second, when the constitutively activated and dominant negative mutants of a small G protein produce the same phenotypes it is generally concluded that the GDP/GTP cycle is necessary as demonstrated for the Arf6-regulated endosomal recycling pathway (27,35,36). However, our results with the fast-cycling Arf6T157N mutant did not fit with this model, as it induced the same phenotypes as the constitutively activated and dominant negative mutants. These two observations could be explained by our model whereby the activated Arf6 would act in cooperation with its GEF.
Furthermore, expression of ARNO did not affect the morphology of the cells, and its C-terminal domain did not show cooperation with Arf6T157N. Thus, by providing a second signal that modulates that of Arf6GTP, EFA6 but not ARNO, specifically regulates the TJ. Hence, one could speculate that depending on the situation, Arf6 could be activated by EFA6 to promote TJ formation or by another Arf6GEF to favor disassembly of E-cadherin-mediated cell-cell adhesion, thereby directing the cells toward either the epithelial-mesenchymal transition or the mesenchymal-epithelial transition pathways. Our findings suggest a new degree in EFA6 specificity. The determination of EFA6 biological functions would not lie solely on its nucleotide exchange activity on Arf6 but would also include the recruitment and co-activation of other partners that contribute to the effects of the activated Arf6. Other Arf6 exchange factors could also be functioning in a similar manner. For example, the effects on the actin cytoskeleton of BRAG2/GEP100 were shown to be dependent on its Sec7 domain and IQ-like motif, suggesting that both regions of the protein might act cooperatively (19). In addition, small interfering RNA depletion of BRAG2 or Arf6 had opposite effects on the cell surface distribution of the ␤1-integrin (20). It might reflect the fact that BRAG2 provides an additional signal to balance Arf6GTP effects. This regulatory mechanism could be a paradigm for other small G proteins and their corresponding GEFs.
Further experiments are required to support our model including the elucidation of the molecular mechanism by which the cooperation of Arf6GTP and its GEF might occur. An important step will be the identification of the direct partners of Arf6 and EFA6 involved in the establishment of the TJ.