Gα12 regulates protein interactions within the MDCK cell tight junction and inhibits tight-junction assembly

The polarized functions of epithelia require an intact tight junction (TJ) to restrict paracellular movement and to separate membrane proteins into specific domains. TJs contain scaffolding, integral membrane and signaling proteins, but the mechanisms that regulate TJs and their assembly are not well defined. Gα12 (GNA12) binds the TJ protein ZO-1 (TJP1), and Gα12 activates Src to increase paracellular permeability via unknown mechanisms. Herein, we identify Src as a component of the TJ and find that recruitment of Hsp90 to activated Gα12 is necessary for signaling. TJ integrity is disrupted by Gα12-stimulated Src phosphorylation of ZO-1 and ZO-2 (TJP2); this phosphorylation leads to dissociation of occludin and claudin 1 from the ZO-1 protein complex. Inhibiting Hsp90 with geldanamycin blocks Gα12-stimulated Src activation and phosphorylation, but does not affect protein levels or the Gα12–ZO-1 interaction. Using the calcium-switch model of TJ assembly and GST-TPR (GST-fused TPR domain of PP5) pull-downs of activated Gα12, we demonstrate that switching to normal calcium medium activates endogenous Gα12 during TJ assembly. Thrombin increases permeability and delays TJ assembly by activating Gα12, but not Gα13, signaling pathways. These findings reveal an important role for Gα12, Src and Hsp90 in regulating the TJ in established epithelia and during TJ assembly.


Introduction
The epithelial cell tight junction (TJ) is the most apical component of the junctional complex. The TJ defines the polarized epithelial phenotype by establishing the barrier to paracellular molecular movement and by providing the fence, restricting the localization of proteins into the apical or basolateral domains. Disruption and reassembly of TJs is an important component of epithelial injury and repair, and contributes to the development of several kidney diseases (Lee et al., 2006). Numerous proteins are found within the TJ, including the transmembrane proteins occludin and junctional adhesion molecule (JAM). However, the claudin family of transmembrane proteins is responsible for the barrier function . In addition, the TJ also contains numerous scaffolding proteins, including the zona-occludens family members ZO-1, ZO-2 and ZO-3 (TJP1, TJP2 and TJP3, respectively), as well as signaling proteins and other unique TJ proteins. There are numerous interactions among TJ proteins, and many of the proteins are linked to the actin cytoskeleton (reviewed in Matter and Balda, 2003). The specific claudins expressed in epithelia are the major determinant of the paracellular properties (Tsukita et al., 2001). The barrier function is also regulated in response to physiological stimuli (reviewed in Nusrat et al., 2000). The finding of numerous signaling proteins within the TJ and the importance of phosphorylation is consistent with extracellular signal regulation of the barrier (see Denker and Nigam, 1998;D'Souza et al., 2005). Furthermore, the biogenesis of the TJ in developing epithelial tissues or after injury involves numerous signaling pathways, which include PKC, Src, small G proteins (Rho and Rac) and G proteins (G␣i2, G␣s) (reviewed in Matter and Balda, 2003).
Heterotrimeric G proteins regulate fundamental cellular properties through their G␣ and G␤␥ subunits (reviewed in Neer, 1995). In addition to signaling from plasma membrane receptors, G␣ subunits are also localized in numerous intracellular compartments and have been implicated in other cellular functions, including vesicular trafficking (Pimplikar and Simons, 1993), protein processing through the Golgi (Stow et al., 1991) and cytokinesis (Blumer et al., 2002). The G␣12/13 family regulates stress-fiber formation and other crucial cellular functions via numerous pathways. Several groups have described G␣ subunits (including G␣12; GNA12) localized to the epithelial cell TJ (de Almeida et al., 1994;Dodane and Kachar, 1996;Ercolani et al., 1990;Hamilton and Nathanson, 1997) and implicated G proteins in regulation of barrier function and TJ assembly (Balda et al., 1991). In the calcium-switch model of TJ assembly (Cereijido et al., 1993;Rodriguez-Boulan and Nelson, 1989), we demonstrated that G␣o, G␣i2 and G␣s stimulate TJ assembly, but only G␣i2 affected baseline TJ properties (Denker et al., 1996;Saha et al., 1998;Saha et al., 2001). We have previously shown that G␣12 and ZO-1 directly interact, and activation of G␣12 leads to TJ disruption and increased permeability (Meyer et al., 2002). The ZO-1-binding site for G␣12 was identified and, upon G␣12 activation, Src but not Rho is activated, leading to disruption of the TJ (Meyer et al., 2003;Meyer et al., 2002). However, the mechanisms regulating the TJ via G␣12 are not known and a potential role for G␣12 in regulating The polarized functions of epithelia require an intact tight junction (TJ) to restrict paracellular movement and to separate membrane proteins into specific domains. TJs contain scaffolding, integral membrane and signaling proteins, but the mechanisms that regulate TJs and their assembly are not well defined. G␣12 (GNA12) binds the TJ protein ZO-1 (TJP1), and G␣12 activates Src to increase paracellular permeability via unknown mechanisms. Herein, we identify Src as a component of the TJ and find that recruitment of Hsp90 to activated G␣12 is necessary for signaling. TJ integrity is disrupted by G␣12stimulated Src phosphorylation of ZO-1 and ZO-2 (TJP2); this phosphorylation leads to dissociation of occludin and claudin 1 from the ZO-1 protein complex. Inhibiting Hsp90 with geldanamycin blocks G␣12-stimulated Src activation and phosphorylation, but does not affect protein levels or the G␣12-ZO-1 interaction. Using the calcium-switch model of TJ assembly and GST-TPR (GST-fused TPR domain of PP5) pulldowns of activated G␣12, we demonstrate that switching to normal calcium medium activates endogenous G␣12 during TJ assembly. Thrombin increases permeability and delays TJ assembly by activating G␣12, but not G␣13, signaling pathways. These findings reveal an important role for G␣12, Src and Hsp90 in regulating the TJ in established epithelia and during TJ assembly. G␣12 regulation of tight junctions TJ assembly has not been investigated. Herein, it is shown that G␣12 activates Src via a mechanism that uses the heat shock protein (Hsp)90 chaperone. Hsp90 interacts with activated [thrombinstimulated or constitutively active QL (glutamine to lysine substitution)] G␣12, but not to inactive G␣12. Src phosphorylates ZO-1 and ZO-2, resulting in loss of claudin 1 and occludin interactions with ZO-1, and inhibition of Hsp90 with geldanamycin blocks these events and prevents barrier disruption. Furthermore, we show for the first time that endogenous G␣12 is activated during TJ assembly in the absence of receptor agonists, and that thrombin stimulation of endogenous G␣12, but not G␣13, increases permeability in intact monolayers and delays TJ assembly.

G␣12 expression in MDCK cells does not affect levels of TJ proteins
Tet-off-inducible wild-type G␣12 and QL␣12 (constitutively active mutant, Q229L) were expressed in Madin-Darby canine kidney (MDCK) cells and have been previously characterized (Meyer et al., 2003;Meyer et al., 2002). Supplementary material Fig. S1 shows the inducible expression of wild-type G␣12 and QL␣12 in MDCK cells with the removal of doxycycline (-dox) for 72 hours; there were no differences in the expression levels of TJ proteins.

Src is a component of the TJ complex and regulates permeability
We previously demonstrated Src-dependent changes in permeability and TJ structure in QL␣12-MDCK cells (Meyer et al., 2003). To further characterize the G␣12-stimulated loss of barrier function, G␣12-and QL␣12-MDCK cells [with/without (+/-) dox] were analyzed for transepithelial resistance (TER, a measure of monolayer ionic permeability), and Na + and Clpermeability. Supplementary material Fig. S2A confirms a small reduction in TER in G␣12-MDCK cells and a significant reduction in QL␣12-MDCK cells. The QL␣12-stimulated reduction in TER is nearly completely inhibited with the Src-specific inhibitor PP2. Consistent with TER data, inhibition of Src with PP2 completely blocked the permeability changes induced with QL␣12 expression (supplementary material Fig. S2B,C, and discussed in more detail later).
To determine whether Src is part of the TJ complex, ZO-1 was immunoprecipitated from G␣12-MDCK cells and probed by western blotting for Src. Fig. 1A shows that Src immunoprecipitates with ZO-1 in QL␣12-MDCK cells independently of G␣12 expression (+/-dox). Controls with rat ascites or beads alone failed to precipitate Src. A similar experiment was repeated in G␣12-and QL␣12-MDCK cells (+/-dox), and the ZO-1 immunoprecipitates were probed by western blotting for active Src (anti-pTyr419) and ZO-1 (Fig. 1B). The amount of inactive Src that co-precipitated with ZO-1 was similar in the four conditions (not shown), but in both G␣12-and QL␣12-MDCK cells there was increased pTyr419 detected in the ZO-1 precipitates of -dox cells. This is consistent with activated Src interacting with the ZO-1 protein complex in both G␣12-and QL␣12-expressing MDCK cells.
To determine colocalization of ZO-1 and Src in MDCK cells, G␣12-MDCK cells (+/-dox) were analyzed by confocal microscopy. Fig.  1C, panels a (+dox) and h (-dox) show a merged image of ZO-1 and Src. To determine the extent of overlap for these proteins, a single cell was identified (Fig. 1C, white box in panels a,h). ZO-1 (Fig. 1C, green), Src (Fig. 1C, red) and merged images are shown at high magnification in the x-y orientation  and i-k (-dox)]. Next, one membrane region was highlighted (Fig. 1C, white enclosed region) and examined in the x-z orientation [Fig. 1C, and l-n (-dox)]. In the absence of G␣12 expression (+dox), ZO-1 was found in a discrete apical region (Fig. 1Ce) and Src was detected in a broader domain of the lateral membrane ( Fig. 1Cf) with some overlap in the apical domain (Fig.  1Cg). With G␣12 expression (-dox), ZO-1 was more widely distributed within the lateral membrane ( Fig. 1Cl-n). This suggests that activation of Src with G␣12 expression (see also Fig. 1B, Fig.  2D and Fig. 3C) is sufficient to induce subtle changes in ZO-1 localization. Taken together with the co-immunoprecipitation of Src with ZO-1, these results reveal that a fraction of Src colocalizes with ZO-1 in both conditions (+/-dox). for ZO-1 and total Src after ZO-1 immunoprecipitation from QL␣12-MDCK cells in +/-dox conditions. ZO-1 hybridoma was added undiluted to an equivalent amount of cell lysates prepared as described in the Materials and Methods. Negative controls were rat ascites using IgG beads. Arrows indicate the ZO-1 and Src bands. (B) Lysates from G␣12-and QL␣12-MDCK cells (+/-dox) were immunoprecipitated with ZO-1 and probed with antibodies against active Src (pTyr419) and ZO-1. (C) Confocal microscopy of G␣12-MDCK cells +/-dox double stained for ZO-1 and Src as described in the Materials and Methods. (Ca,Ch) White box indicates the cell that was examined at higher magnification in the panels to the right. In x-y and lat-z panels, a region of the membrane was identified (white enclosure) for additional analysis. The lat-z images represent the apical 5 m region of the cell (total height approximately 8 m). (Ca-Cg) +dox; (Ch-Cn) -dox. FITC-ZO-1, Texas-red c-Src and merged images are shown for both conditions in x-y and lat-z orientation. Scale bars: 10 m (Ca,Ch), 5 m (Cb-Cg,Ci-Cn).

QL␣12 stimulates Src tyrosine phosphorylation of ZO-1 and ZO-2
To characterize the tyrosine phosphorylation of TJ proteins, QL␣12-MDCK cells (+/-dox) were cultured with or without PP2 (Src inhibitor) or genistein (inhibits numerous tyrosine kinases). Immunoprecipitations of ZO-1, ZO-2, occludin ( Fig. 2), claudin 1 and claudin 2 (not shown) were followed by phosphotyrosine western blots. In Fig. 2, the first panel shows tyrosine phosphorylation of ZO-1 ( Fig.  2A), ZO-2 ( Fig. 2B) and occludin (Fig. 2C) in +/-dox conditions. With induction of QL␣12, the tyrosine phosphorylation of ZO-1 ( Fig. 2A) and ZO-2 ( Fig. 2B) was significantly increased (two-to three-fold), whereas there was no significant change for occludin (Fig. 2C). No tyrosine phosphorylation of claudin 1 and claudin 2 was detected under any condition (not shown). In G␣12-MDCK cells there was significantly less tyrosine phosphorylation of ZO-1 at baseline (+dox), although a small increase was seen with G␣12 expression (-dox; Fig.  2D). It was difficult to detect any baseline or stimulated tyrosine phosphorylation of ZO-2 or occludin in G␣12-MDCK cells (not shown). Pre-treatment of QL␣12-MDCK cells with either PP2 or genistein inhibited QL␣12stimulated ZO-1 phosphorylation to levels that were similar to the +dox control. The stimulation of ZO-2 tyrosine phosphorylation with QL␣12 expression was threefold higher than in controls and, unlike ZO-1, was completely inhibited with PP2 and genistein to levels significantly below basal. Tyrosine phosphorylation on occludin was not significantly affected by the tyrosine kinase inhibitors.

QL␣12 disrupts occludin and claudin 1 interactions with ZO-1
The epithelial cell TJ in MDCK cells contains a complex of interacting proteins (reviewed in Matter and Balda, 2003). We next asked whether G␣12-Src-stimulated tyrosine phosphorylation of ZO-1 and ZO-2 affected protein-protein interactions within the TJ. ZO-1 was immunoprecipitated and interacting proteins analyzed by western blot. Fig. 3A shows that control immunoprecipitations with rat ascites did not precipitate ZO-1 from QL␣12-MDCK cells +/-dox, whereas 2-4% of total ZO-1 was immunoprecipitated with ZO-1 hybridoma under these conditions. G␣12 expression in MDCK cells (+/-dox) did not affect the amount of immunoprecipitated ZO-1 (Fig. 3B) and reprobing the blot with antibodies to other TJ proteins demonstrated that the interaction between ZO-2 or occludin with the ZO-1 complex was not affected. However, there was a small but consistent reduction in the amount of claudin 1 interacting with the ZO-1 complex (24±7%; n=3) ( Fig. 3B,C). Like in G␣12-MDCK cells, in QL␣12-MDCK cells there were no changes in the amount of ZO-1 precipitated nor of co-precipitated ZO-2 (Fig. 3B). However, there was significant loss of occludin and nearly complete disruption of claudin 1 (Fig. 3B,C; 54±12% and 75±15%, respectively) from the ZO-1 complex. To determine whether Src modulated these protein interactions, the immunoprecipitations were repeated in cells treated with PP2. As shown in Fig. 3B,C, Src inhibition completely prevented the loss of occludin and claudin 1 from the ZO-1 complex. These findings suggest that G␣12/Srcstimulated tyrosine phosphorylation of ZO-1 and ZO-2 leads to disruption of TJ-protein interactions.

Journal of Cell Science 121 (6)
The dissociation of claudin 1 from the ZO-1 complex was further characterized by double labeling and immunofluorescence microscopy. Supplementary material Fig. S3A confirms the previously reported localization of ZO-1 and claudin 1 in +/-dox and in -dox +PP2 QL␣12-MDCK cells (Meyer et al., 2003). In +dox conditions, both ZO-1 and claudin 1 were localized along the basolateral membrane with partial overlap. With QL␣12 expression (-dox) there was a loss of barrier function (supplementary material Fig. S2), and disruption of ZO-1 and claudin 1 staining (supplementary material Fig. S3A). Pre-treatment of QL␣12-MDCK cells with the Src inhibitor PP2 blocked the changes in ZO-1 and claudin 1 staining patterns (supplementary material Fig. S3A). Finally, the redistribution of claudin 1 was examined biochemically with detergent extraction (supplementary material Fig. S3B). Claudin 1 was approximately 51±5% (n=2) soluble in +dox conditions, but became more resistant to detergent extraction with QL␣12 expression (26±7% soluble in -dox conditions). Consistent with the immunoprecipitation and immunofluorescence microscopy Immunoprecipitations of the indicated TJ proteins were performed as described in the Materials and Methods followed by western blotting with 4G10 antibody. Identical amounts of lysate from QL␣12-MDCK cells +/-dox for 72 hours were used for each precipitation. Immunoprecipitations were repeated from cells cultured in PP2 (10 M) or genistein (2 M), which were added to the medium at the time of medium change to -dox. The intensity of the phosphotyrosine (pTyr) band (arrows) in each condition is normalized to the amount of protein in the precipitate. This fraction was divided by the fraction of pTyr detected in +dox cells with no inhibitor condition. Bar graphs on the right summarize the results from three experiments. (A) ZO-1; (B) ZO-2; (C) occludin; and (D) ZO-1 in G␣12-MDCK +/-dox. *P<0.05. G␣12 regulation of tight junctions results, inhibiting Src with PP2 prevented the QL␣12 resistance to detergent solubilization.

Hsp90 modulates QL␣12-mediated increases in paracellular permeability
G␣12 does not directly interact with Src and does not stimulate Src kinase activity (Ma et al., 2000). Hsp90 interacts with both G␣12 (Vaiskunaite et al., 2001;Waheed and Jones, 2002) and Src (Xu and Lindquist, 1993), so we speculated that the Hsp90 chaperone might provide the link between activated G␣12 and Src activation. To test this possibility, we used the Hsp90 inhibitor geldanamycin (GA) (Prodromou et al., 1997). In addition to characterizing the barrier by TER measurement (Fig. 4A), we also measured transepithelial permeability to individual ions using diffusionpotential measurements in Ussing chambers. In un-induced MDCK II cells (+dox), Na + permeability was about ninefold higher than Clpermeability, indicating the presence of a strong paracellular barrier to anions. Induction of QL␣12 expression (-dox) abolished the normal cation selectivity, primarily because of a marked increase in Clpermeability (Fig. 4B). In +dox cells, the normal alkali metal selectivity sequence (Li~Na>K>RbӷCs) conformed to Eisenman sequence X-XI, suggesting that cation permeation is dependent on electrostatic binding to a strong, negatively charged site within the pore (Diamond and Wright, 1969). Induction of QL␣12 almost completely reversed this selectivity (Cs~Rb~K>Na>Li). The resulting sequence (Eisenman I-II) was the same as the order of were cultured +/-dox +/-PP2 and immunoprecipitated with rat monoclonal R46 antibody or the equivalent amount of ascites control. Western blot for ZO-1 in cell lysates and immunoprecipitations using the identical antibody is shown. (B) G␣12-and QL␣-12 MDCK cells +/-dox were immunoprecipitated with ZO-1 antibody as described in the Materials and Methods. Immunoprecipitates were analyzed by western blotting to ZO-1 followed by reprobing with antibodies to other TJ proteins (ZO-2, occludin and claudin 1). A parallel analysis was performed in QL␣12-MDCK cells pre-treated with the Src inhibitor PP2 (10 M) added at the time of switch to -dox medium. (C) Bar graphs summarizing the fraction of TJ protein remaining associated with ZO-1. The results are the mean ± s.e.m. of three experiments normalized to the ZO-1 immunoprecipitates (+dox). *P<0.05. free solution mobility of the hydrated ions, suggesting that, under these conditions, cation transport occurred by aqueous diffusion. Finally, induction of QL␣12 impaired the size-selective barrier. The normal threefold selectivity of Na + (diameter 1.9 Å) over that of arginine, a large organic cation (diameter 7 Å), was completely abolished (Fig. 4C) in QL␣12-expressing MDCK cells. Fig. 4 shows TER, sodium and chloride permeability, and size selectivity in QL␣12-MDCK cells treated with GA. For all parameters tested, GA had no effect in +dox conditions (Fig. 4). The changes in TER, sodium and chloride permeability, and size selectivity induced with QL␣12 expression (-dox) were all reversed with Hsp90 inhibition by GA ( Fig. 4A-C). This was identical to the findings with the Src inhibitor PP2 (supplementary material Fig. S2B,C). We next determined whether the morphological changes in the TJ induced by QL␣12 expression were reversed by Hsp90 inhibition. Fig. 4D shows ZO-1, occludin and claudin 1 staining of the lateral membrane in control (+dox) QL␣12-MDCK cells. With QL␣12 expression (-dox), ZO-1 and occludin staining became irregular and diffuse along the plasma membrane, and claudin 1 was predominantly intracellular (Fig. 4D). GA treatment prevented the dislocation of ZO-1, occludin and claudin 1 from the lateral membrane (Fig. 4D, last panel). Finally, inhibition of Hsp90 with GA also prevented the changes in claudin 1 solubility (supplementary material Fig. S3B).

Inhibition of Hsp90 blocks QL␣12-stimulated Src activation
Src activity was assessed in QL␣12-MDCK cells cultured in +/-dox and +/-GA. Fig. 5A shows the results of western blotting to Src pTyr 419, an indicator of relative Src activity. As previously demonstrated (Meyer et al., 2003), there is a 50% increase in pTyr419 immunoreactivity in -dox QL␣12 cells (compared with +dox; Fig. 5A,B). However, inhibiting Hsp90 with GA blocked the QL␣12-stimulated increase in Src activity without affecting total Src or occludin levels (Fig. 5A). We expected that GA would inhibit tyrosine phosphorylation of TJ proteins. Because the predominant G␣12-Src target in the TJ is ZO-2 ( Fig. 2), we immunoprecipitated ZO-2 from QL␣12-MDCK cells +/-dox and +/-GA followed by western blotting for phosphotyrosine (Fig. 5C). Inhibition of Hsp90 blocked G␣12/Src-stimulated phosphorylation of ZO-2. The effects of GA on ZO-1 phosphorylation were inconsistent and it is unclear whether ZO-1 phosphorylation is affected by Hsp90 inhibition (not shown). QL␣12 interacts with ZO-1 and can be immunoprecipitated with the ZO-1 complex.
To determine whether inhibiting Hsp90 affects this interaction, ZO-1 was immunoprecipitated from QL␣12-MDCK cells +/-dox and +/-GA. Fig. 5D shows that, in the absence of QL␣12 expression, no QL␣12 was detected. Immunoprecipitation of ZO-1 from QL␣12-MDCK (-dox) cells co-precipitated a fraction of QL␣12. Treatment with GA did not disrupt the QL␣12-ZO-1 complex but rather appeared to enhance it; there was a small increase in the amount of coprecipitated QL␣12 (n=4, Fig. 5D). The finding suggests that Hsp90 is not required for the G␣12-ZO-1 interaction and is consistent with a model in which Hsp90 is recruited to the complex (discussed later).

Journal of Cell Science 121 (6)
The steady-state levels of QL␣12 and TJ proteins +/-dox and +/-GA in QL␣12-MDCK cells were assessed and no significant differences were seen (Fig. 5E). The Hsp90 chaperone complex includes numerous co-chaperones (reviewed in Whitesell and Lindquist, 2005) and a novel N-terminal myristoylated Hsp90 isoform (Hsp90N) is expressed in epithelial cells (Grammatikakis Fig. 5. Hsp90 is a component of the activated G␣12 complex and is necessary for Src activation and ZO-2 phosphorylation. (A) Src activation in QL␣12-MDCK cells +/-dox and +/-GA. Identical amounts of cell lysate were analyzed by western blotting with antibodies specific for pTyr-419 as described in the Materials and Methods. Membranes were stripped and reanalyzed for total Src and occludin. (B) Band intensity was quantified as described above and the pTyr-419 immunoreactivity normalized to total Src. The relative amount of pTyr419 for -dox and -dox +GA is shown in comparison to the +dox condition. (C) Phosphotyrosine content of immunoprecipitated ZO-2 from QL␣12-MDCK cells +/-dox and +/-GA. ZO-2 was immunoprecipitated and analyzed with 4G10 anti-phosphotyrosine as described above. (D) Interaction of G␣12 with ZO-1 in QL␣12-MDCK cells +/-dox and +/-GA. ZO-1 was immunoprecipitated under the specified conditions and precipitates were analyzed for G␣12 by western blotting. Control is lysate with beads alone. (E) Steady-state protein levels of TJ proteins and G␣12 in QL␣12-MDCK +/-dox and +/-GA (2 M). (F) Validation of GST-TPR pulldown of activated G␣12. QL␣12-and G␣12-MDCK cells were cultured in -dox, and lysates (~800 g) were pulled-down with 1 g GST or GST-TPR, as described in the Materials and Methods, and analyzed by western blotting for G␣12. 10% of the input of wild-type G␣12 is shown in the lysate lane. (G) Hsp90 western blot of GST-TPR pulldowns from G␣12-and QL␣12-MDCK cells in -dox. G␣12 regulation of tight junctions et al., 2002). However, we were unsuccessful in detecting these proteins in ZO-1 immunoprecipitates under a variety of conditions (not shown). Recently, a GST pull-down assay using GST-TPR [tetratricopeptide domain of protein phosphatase 5 (PP5)] for detecting activated G␣12 has been reported (Yamaguchi et al., 2002;Yamaguchi et al., 2003). This domain selectively interacts with activated G␣12 and G␣13, and has been validated in our G␣12-MDCK cell models (Yanamadala et al., 2007). We reconfirmed the assay using lysates from G␣12-and QL␣12expressing MDCK cells (-dox; Fig. 5F) and studied whether Hsp90 might be a component of the activated G␣12 complex. Fig.  5G shows that Hsp90 is detected in pull-downs of activated (QL) but not wild-type G␣12. This observation is consistent with Hsp90 preferentially interacting with active G␣12. This finding was confirmed and extended with thrombin activation of endogenous G␣12 (discussed later).

A role for G␣12 signaling in regulating TJ assembly
We previously demonstrated that activated G␣i2 and G␣s stimulated TJ assembly in the calcium switch and speculated that there were counter-regulatory G␣mediated pathways (Saha et al., 1998;Saha et al., 2001). G␣12 was a good candidate to provide the 'brake' on TJ assembly. To address this possibility, we stimulated control MDCK cells (G␣12 in +dox) with thrombin, a known agonist for G␣12/G␣13-coupled pathways. Fig.  6A shows that the addition of thrombin at t=0 significantly delays TJ assembly. The GST-TPR pull-down of activated G␣12 (Fig. 6B) revealed that the addition of thrombin leads to robust activation of endogenous G␣12 by 30 minutes and the activation persists through to 2 hours. As expected, endogenous G␣12 was below the detection limit in G␣12-MDCK cells in +dox (lysate lane in Fig.  6B) but, surprising, we detected activation of endogenous G␣12 in non-stimulated cells at 1-2 hours after inducing TJ assembly. We next examined the effects of thrombin on TJ assembly in the calcium switch using G␣12-MDCK cells in -dox. Fig. 6C shows that the effects of thrombin on TJ assembly in G␣12-expressing MDCK cells are similar to the control (+dox; Fig. 6A). Thrombin stimulation resulted in prolonged G␣12 activation (at least 8 hours) in -dox cells (Fig. 6D). Again, there was readily detectable activation of G␣12 without thrombin stimulation seen in GST-TPR pull-downs at approximately 1 hour after the calcium switch. These findings indicate that endogenous G␣12 is activated in non-stimulated MDCK cells undergoing TJ assembly, and that thrombin significantly delays TJ assembly and robustly activates G␣12.
However, thrombin also activates G␣13 and G␣q pathways, and these results do not exclude a role for other G-protein-coupled pathways. The GST-TPR pull-down experiments were probed for G␣13, and Fig. 6E shows that there was no detectable activation of G␣13 in the absence of thrombin stimulation in either +/-dox cells. With thrombin stimulation, there was activation of G␣13 at 30 minutes that was no longer detectable by 1-2 hours after the calcium switch. To determine whether thrombin might regulate TJ assembly through G␣13 or G␣q pathways, the calcium-switch experiments were repeated +/-bradykinin, an agonist reported to activate G␣13, G␣q and other G␣ subunits, but not G␣12 (reviewed in Riobo and Manning, 2005). Fig. 7A,B shows that bradykinin stimulation of MDCK cells +/-dox during the calcium switch had no effect on the time-course of TJ assembly. Analysis of GST-TPR pull-downs from bradykinin-stimulated cells confirmed G␣13 activation from 30 minutes to 2 hours with little or no additional activation of G␣12 (Fig. 7C). These results indicate that bradykinin activation of G␣13 and other G␣ pathways is unlikely to affect TJ assembly, and are consistent with thrombin regulating TJ assembly through G␣12.
We next investigated whether the inhibition of TJ assembly stimulated by thrombin-G␣12 used Src and Hsp90. The calciumswitch experiments were repeated in control cells (+dox) +/-thrombin and +/-the Src inhibitor PP2 (Fig. 8A) or the Hsp90 inhibitor GA (Fig. 8B). The addition of either inhibitor without thrombin had no effect on the time-course of TER recovery in the calcium switch (Fig. 8A,B). When thrombin was added in the presence of PP2 there was partial recovery in the assembly of the Fig. 6. Endogenous G␣12 is activated during TJ assembly in the calcium switch and thrombin delays TJ assembly. (A) Calcium switch in control G␣12-MDCK cells +dox. Cells were cultured overnight in low calcium (LC; t=0) medium and then switched back to normal calcium (NC) medium. TER measurements were obtained at the specified intervals. Thrombin (2 U/ml) was added at t=0. Results are the mean ± s.e.m. for n=4 at each time point. The experiment was repeated three times with nearly identical results. (B) Western blot of G␣12 +dox cells +/-thrombin analyzed by GST-TPR pull-down at the specified times after the calcium switch. (C) Calcium switch in G␣12-MDCK cells with induced G␣12 expression (-dox). (D) Western blot of G␣12 -dox cells +/-thrombin analyzed by GST-TPR pull-down at the specified times after the calcium switch. The first lane of the top panel shows the induced G␣12 and is ~10% of the protein used for the pull-downs. (E) Western blots for G␣13 on G␣12-MDCK cells +/-dox and +/-thrombin and pulled-down with GST-TPR as described in C. For each panel, the +/-thrombin results were obtained from a single gel. barrier (Fig. 8A) and GA completely reversed the thrombinstimulated delay in barrier assembly (Fig. 8B). These observations suggest that TJ regulation by G␣12 in intact monolayers and during TJ assembly use similar mechanisms.
Finally, we studied whether thrombin is a physiologic regulator of epithelial cell TJs. Fig. 9A shows that the addition of thrombin to a confluent monolayer of control MDCK cells leads to the rapid loss of TER (within minutes). The effects of thrombin were maximal at 1-2 hours and persisted for several more hours before gradually returning to control levels. To determine whether Src and Hsp90 were important to the thrombin signaling, TER values were compared at 40 minutes after the addition of thrombin, bradykinin, Journal of Cell Science 121 (6) PP2 and GA (alone and in combination, Fig. 9B). PP2 and GA had no independent effect on TER (Fig. 9B) and, consistent with bradykinin activation of G␣13 but not G␣12, there were no effects of bradykinin on baseline TER (Fig. 9B). Inhibition of Src with PP2 or inhibition of Hsp90 with GA led to normalization of the thrombin-stimulated decrease in TER (Fig. 9B). To confirm the time-course of thrombin stimulation, G␣12 activation and recruitment of Hsp90 to the activated complex are shown in GST-TPR pull-downs at t=0, 15 and 30 minutes after the addition of thrombin (Fig. 9C). Neither protein was seen at t=0 but, within 15 minutes of thrombin stimulation, both G␣12 and Hsp90 were detected, and at 30 minutes levels of both proteins were increased. Taken together, these findings suggest that thrombin might physiologically regulate TJs via G␣12 activation, recruitment of Hsp90 and Src activation.

Discussion
These studies reveal how G␣12 regulates the TJ in established epithelia and provide the first evidence that G␣12 negatively regulates TJ assembly. It was previously shown that G␣12 interacts with ZO-1, and activated G␣12 disrupts TJ structure and function through Src tyrosine kinases (Meyer et al., 2003;Meyer et al., 2002). Now, it is shown that this mechanism involves disrupting proteinprotein interactions within the multi-protein TJ complex. This occurs via Src phosphorylation of ZO-1 and ZO-2, leading to displacement of occludin and claudin 1 from the ZO-1 protein complex. We identified Hsp90 as the intermediate that links activated G␣12 to Src activation. Surprisingly, we identified activation of endogenous G␣12 in the calcium-switch model of TJ biogenesis in the absence of any receptor agonists, suggesting that it might be a crucial checkpoint in TJ assembly. Stimulation of MDCK cells with physiological concentrations of thrombin activates G␣12 but not G␣13, and leads to increased paracellular permeability and delayed TJ assembly through pathways that involve Src and the Hsp90 chaperone complex.  Numerous studies have found that tyrosine phosphorylation of TJ proteins perturbs the barrier (Collares-Buzato et al., 1998), and tyrosine kinase phosphorylation of TJ proteins has been shown to be important in the assembly of the junction in models of ATP depletion and reactive oxygen species (Meyer et al., 2001;Tsukamoto and Nigam, 1999). Src mediates oxidative-stressinduced disruption of the TJ in Caco-2 cells (Basuroy et al., 2003) and, in an injury model of mouse proximal tubular cells, Src was localized to the junctions (Sinha et al., 2003). However, there has been little direct evidence of Src within the junctional complex.
Our results indicate that a fraction of total Src is found in the TJ complex, and is likely to account for the tyrosine phosphorylation of ZO-1, ZO-2 and possibly other proteins within this membrane microdomain. Nevertheless, the results with genistein and the absence of changes in occludin phosphorylation indicate that additional tyrosine kinases are also important in TJ regulation. The findings in this study are consistent with earlier studies in temperature-sensitive-v-Src-transfected MDCK cells, in which Src phosphorylated ZO-1 and ZO-2 and led to increased permeability (Takeda and Tsukita, 1995). Although ZO-1, ZO-2 and ZO-3 contain an SH3-binding domain, our immunoprecipitation studies could not distinguish direct versus indirect interactions between Src and ZO-1. Although Src could be detected within the ZO-1 complex at baseline, activation of G␣12 led to an increase in the fraction of activated Src interacting with ZO-1 (Fig. 1B). The SH3 domain of ZO-1 is especially important for TJ structure and function; we identified this region of ZO-1 as the interaction site with G␣12 (Meyer et al., 2002) and the SH3 domain alone was shown to rescue delayed TJ assembly in ZO-1-silenced MDCK cells (McNeil et al., 2006). One important role for ZO-1 and related proteins might be to organize and retain a complex of signaling proteins within the TJ in order to provide local regulation in response to changes in the cell environment. The interactions of G␣12 with ZO-1 and the discovery of Src within this complex are consistent with this possibility. We initially identified G␣12 interacting with ZO-1 and ZO-2 (Meyer et al., 2002), and now find that ZO-2 tyrosine phosphorylation was more pronounced than with ZO-1. This finding suggests that, in vivo, the more relevant interaction of G␣12 might be with ZO-2. Future studies will be needed to distinguish between these possibilities.
Our electrophysiological studies suggest that activation of G␣12 induces loss of all aspects of paracellular selectivity. Normally in MDCK cells there is a strong paracellular charge selectivity favoring permeation of cations over anions. Among alkali metal cations, there is selectivity favoring the most heavily hydrated ions (Cereijido et al., 1978;Van Itallie et al., 2001), suggestive of a strong electrostatic cation-binding site in the pore (Diamond and Wright, 1969). Finally, there is normally marked size selectivity favoring permeation of small inorganic cations over bulky organic cations, such as arginine (Yu et al., 2003). We found that QL␣12 abolishes all three types of selectivity, suggesting that it induces the opening of large non-selective hydrated pores. This seems unlikely to be due to subtle changes in the expression of specific claudin isoforms and would be more consistent with mechanical disruption of the paracellular barrier by perturbation of the scaffolding of the TJ. Our finding that QL␣12 disrupts the interaction between ZO-1 and claudin 1 would be most consistent with this possibility. The implications for the loss of occludin from the ZO-1 complex with G␣12 activation are less clear. Although occludin is an integral membrane protein, it is not required for the formation of an intact TJ (Saitou et al., 1998).
We hypothesized that heat-shock proteins were likely to mediate G␣12-Src signaling. Hsp90 has been widely implicated in stabilizing signaling proteins, including G␣ subunits (Busconi et al., 2000) and Src (Xu and Lindquist, 1993;Xu et al., 1999), and Hsp90 directly interacts with G␣12 (but not G␣13) (Vaiskunaite et al., 2001;Waheed and Jones, 2002). G␣12 also activates tyrosine kinases, including Src, but not directly (Jiang et al., 1998;Ma et al., 2000;Nagao et al., 1999). GA treatment of MDCK cells completely inhibited G␣12-stimulated signaling and blocked Src activation (Fig.  5A,B). Inhibition of Hsp90 with GA did not affect the interaction of G␣12 with ZO-1 (Fig. 5D) and we were able to demonstrate Hsp90 in the complex only after G␣12 activation ( Fig. 5G and Fig.  9). Other reports support the finding that heat-shock proteins can interact with the TJ; in endothelial cells, an interaction between Hsp70 and ZO-1 has been reported (Lu et al., 2004), and another member of the Hsp70 family, Apg-2, binds to ZO-1, localizes in the TJ and co-precipitates after heat shock (Tsapara et al., 2006). Taken together, we speculate that ZO-1, Src, G␣12 and other components of the TJ complex are in close proximity and/or in Fig. 9. Thrombin stimulates the loss of barrier function in intact monolayers through G␣12, Src and Hsp90. (A) Confluent monolayers of G␣12-MDCK cells in +dox on Transwell filters were stimulated with thrombin (2 U/ml) at t=0 and TER was measured over 12 hours. Results are the mean ± s.e.m. of n=4 and were normalized to the untreated control. (B) Monolayers were stimulated at t=0 with thrombin (2 U/ml), PP2 (10 M), GA (0.2 M), thrombin plus PP2 or GA, bradykinin (100 M), or bradykinin plus PP2 or GA and compared to vehicle (control). TER was measured at numerous time points. The results (normalized to the untreated control) at 40 minutes are shown (n=4 ± s.e.m.). *P<0.05. (C) Confluent G␣12-MDCK cells in +dox were stimulated at t=0 with thrombin and lysates were prepared at t=0 (no thrombin), 15 and 30 minutes of stimulation for analysis by GST-TPR pulldown. Western blots to G␣12 and Hsp90 were performed as described above.
Journal of Cell Science 121 (6) direct contact with each other (see Fig. 10). With activation of G␣12 (QL mutation or thrombin), there is recruitment of Hsp90 to the G␣12-ZO-1 protein complex, facilitating activation of Src (Fig. 10). It is likely that only a small fraction of total cellular G␣12 and Src would be required for signaling in this domain, and it is the association of G␣12 and Src with ZO-1 (or other scaffolding proteins in the junction) that maintains the local organization of the signaling complex.
The mechanisms that epithelia use to assemble a competent TJ involve numerous signaling pathways. The calcium switch is widely used as a model of TJ biogenesis and recapitulates many of the crucial molecular events in epithelial morphogenesis (Balda et al., 1991;Cereijido et al., 1993;Rodriguez-Boulan and Nelson, 1989). The signaling events leading to TJ assembly in the calcium switch use numerous pathways (reviewed in Matter and Balda, 2003). We previously identified roles for several G␣ subunits in regulating the assembly of the TJ in the calcium switch (Denker et al., 1996;Saha et al., 1998;Saha et al., 2001) but in each case the kinetics of TJ formation were accelerated. We hypothesized that other G␣ subunits were likely to counter-regulate this process and now identify G␣12 as a crucial check-point negatively regulating TJ assembly. The mechanism(s) leading to activation of G␣12 during this process are currently being investigated and could involve activation of calcium-sensing receptors (CaR) that couple to G␣12 in MDCK cells . Thrombin robustly activates G␣12, significantly delays TJ assembly and affects baseline properties within minutes. The physiological importance of thrombin regulation of epithelial cell permeability is supported by studies of intestinal PAR2 receptors; in these studies, thrombin led to increased permeability and redistribution of ZO-1 and occludin from the TJ (Jacob et al., 2005). Other G␣12/G␣13 receptor agonists such as LPA also affect barrier function (Postma et al., 1998;Schulze et al., 1997).
Taken together, these results suggest that there is a balance between signaling pathways favoring either TJ assembly or disassembly. G␣12/Src signaling is likely to be a pivotal mechanism in the pathways leading to increased permeability and TJ disassembly (Fig. 10). This possibility is supported by the observation that similar signaling mechanisms, via G␣12, Hsp90 and Src, are important for regulating paracellular permeability in established monolayers and for slowing TJ assembly. We propose that G␣12 activation is important in the normal regulation of TJ function and in pathophysiological conditions that disrupt TJs (such as ischemic renal injury). Insights from our studies suggest that inhibiting G␣12 and Src might permit more rapid recovery of epithelial cell function after injury.

Transepithelial resistance
MDCK cells were plated at confluency on polycarbonate Transwell filters (Transwell, Costar) at ~4ϫ10 5 cells/cm 2 and maintained in culture medium +dox for 72 hours. Monolayers were maintained in +dox medium or switched to -dox for 72 hours in the presence and absence of inhibitors, and TER was measured using the Millicell Electrical Resistance system (Millipore, Bedford, MA) as previously described (Denker et al., 1996). Measurements (in Ω/cm 2 ) are expressed as a mean of the original readings (n=3 to 5) after subtraction of background values.

Diffusion potential measurements
Measurements were performed as described previously (Yu et al., 2003). In brief, cells cultured on Snapwell filters (Corning) were mounted in Ussing chambers that were filled with Ringer solution (150 mM NaCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM glucose, 10 mM HEPES, pH 7.4) at 37°C. To determine the NaCl dilution potential, the solution in the basolateral chamber was replaced by a modified Ringer solution containing 75 mM NaCl (osmolarity adjusted with mannitol) and the voltage was read immediately under open-circuit conditions. The basolateral solution was then sequentially replaced by Ringer solution in which NaCl was substituted by 150 mM LiCl, KCl, RbCl, CsCl and arginine chloride, to measure the bi-ionic potential. Relative permeability ratios were calculated from these diffusion-potential data using the Goldman-Hodgkin-Katz equation. The absolute permeabilities to Na + and Clwere derived by the method of Kimizuka and Koketsu (Kimizuka and Koketsu, 1964).

Immunoprecipitation and western blotting
G␣12-or QL␣12-expressing MDCK cells were cultured +/-dox. Monolayers were scraped in lysis buffer [100 mM NaCl, 2 mM EDTA, 10 mM HEPES, pH 7.5, 1 mM NaVO 4 , 25 mM NaF, 1 mM phenylmethylsulfonylfluoride (PMSF), 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM PMSF and proteases inhibitors (Roche, Indianapolis, IN)], frozen/thawed, triturated and centrifuged. Immunoprecipitations of ZO-1 (with R40.76), ZO-2, occludin, claudin 1 and claudin Fig. 10. G␣12 as a check-point in TJ regulation. At baseline (A), claudins from neighboring cells interact with each other in the paracellular space and with ZO-1 in the TJ. The TJ protein complex contains numerous proteins. In this figure, ZO-1, ZO-2 and ZO-3 are designated ZO. Src might interact directly or indirectly with the ZO complex. G␣12 directly interacts with ZO-1 and ZO-2, and Hsp90 is probably not constitutively localized in this domain but is recruited to the activated G␣12 complex (denoted with * in B). Activated G␣12 (QL mutation or thrombin) activates Src (denoted with * in B) leading to increased tyrosine phosphorylation of ZO-1 and ZO-2 (P-Y). (C) Phosphorylation of ZO-1 and ZO-2 leads to disruption of the ZO-1 protein complex, intracellular localization of claudin 1 and loss of barrier function. In addition to the loss of barrier function, under some conditions of persistent activation, this could lead to persistent disassembly of the TJ. Once TJs are disrupted and stimulated to reassemble (as in the calcium switch), activation of G␣12 (using Hsp90 and Src) leads to a brake on TJ assembly. G␣12 regulation of tight junctions 2 were done with protein-A-sepharose or goat affinity-purified rat IgG beads (MP Biomedicals, Aurora, OH) overnight at 4°C. Beads were centrifuged, washed three times and proteins eluted in SDS-PAGE sample buffer. SDS-PAGE was followed by western blot transfer as previously described (Meyer et al., 2003). After blocking and washing, primary antibodies to ZO-1 (rat 1:10 or rabbit 1:100), ZO-2 (1:1000), occludin (1:1000), claudin 1 (1:750) and claudin 2 (1:750) were added overnight followed by incubation with secondary horseradish-peroxidase-conjugated antibodies for 1 hour. Signal was detected with SuperSignal West Pico horseradish peroxidase substrate system (Pierce) and autoradiography (Biomax MR, Eastman Kodak, Rochester, NY).

Detergent extractions
Detergent-soluble and -insoluble fractions were prepared as described previously (Singh and Harris, 2004). Extractions were done at 72 hours after switching to -dox using extraction buffer (10 mM HEPES, pH 7.2, 1% triton X-100, 100 mM NaCl and protease inhibitors). Cells were scraped, incubated for 20 minutes at 4°C and centrifuged. The supernatant (S; soluble fraction) was collected and the pellet resuspended in solubilization buffer (10 mM HEPES, pH 7.2, 1% SDS, 100 mM NaCl and protease inhibitors). The cell suspension was sonicated, incubated for 20 minutes at 4°C and centrifuged for 20 minutes. The resulting supernatant was labeled I (insoluble fraction). Equivalent volumes of each fraction were analyzed by SDS-PAGE and western blotting.
Immunohistochemistry and confocal imaging G␣12-and QL␣12 MDCK cells were plated on Transwell filters and maintained +/-dox for 72 hours. The cells were fixed with 4% paraformaldehyde (EM Sciences, Fort Washington, PA) in PBS for 10 minutes at room temperature and rinsed with PBS three times for 5 minutes. The cells were blocked for 45 minutes at room temperature in PBS-Triton X 0.02% and 5% (w/v) BSA (PBS-T). Primary antibodies were diluted 1:200 in PBS-T and incubated at room temperature for 1.5 hours. Texas red or fluorescein-isothiocyanate-conjugated secondary antibodies (Pierce, Rockford, IL) were diluted 1:100 in 0.05% Triton X-100 in PBS. Filters were washed three times with PBS for 5 minutes, mounted, and imaging was carried out using a Nikon TE 2000E2 microscope, captured to a COOLSNAP HQ cooled CCD (Photometrics; Tuscon, AZ) and IPLab software (Scanalytics; Fairfax, VA). Confocal imaging was carried out using a CSU21 UltraView Spinning Disk Confocal Imager (Perkin Elmer; Boston, MA) and three-dimensional reconstruction performed in the Volocity software package (Improvision; Lexington, MA). Cells were imaged from the apical region to within a few microns of the basolateral surface to reduce autofluorescence originating from the basolateral membrane filter. Specific images were assembled using Photoshop software (Adobe; San Jose, CA) and incorporated into figures using Illustrator (Adobe; San Jose, CA).

GST-TPR pull-down
GST-TPR was purified from bacterial lysates as described (Meyer et al., 2002). MDCK cells were cultured under specified conditions and cell lysates prepared in lysis buffer (50 mM HEPES, pH 7.5, 1 mM EDTA, 3 mM DTT, 10 mM MgSO 4 , 1.0% polyoxyethylene-10-lauryl ether and protease inhibitors). 1 g of GST-TPR or GST alone were coupled to glutathione-agarose beads (Amersham Pharmacia), added to 800 g of total protein and rocked overnight at 4°C. Samples were centrifuged, the beads were washed three times with PBS+0.1% Triton X-100 and resuspended in lysis buffer containing SDS-PAGE sample buffer. Eluted samples were analyzed by SDS-PAGE and western blotting.

Quantification and statistics
Western blots were scanned using an Epson 1640 desktop scanner and band intensity quantified using NIH Image (Wayne Rasband) after subtracting background and determining linear range. Statistics were done in GraphPad Prism (San Diego, CA). Significance was determined by using two-tailed t-test.