Zonula occludens-1 is a scaffolding protein for signaling molecules. Galpha(12) directly binds to the Src homology 3 domain and regulates paracellular permeability in epithelial cells.

Zonula occludens proteins are multidomain proteins usually localized at sites of intercellular junctions, yet little is known about their role in regulating junctional properties. Multiple signaling proteins regulate the junctional complex, and several (including G proteins) have been co-localized with zonula occludens-1 (ZO-1) in the tight junction of epithelial cells. However, evidence for direct interactions between signaling proteins and tight junction proteins has been lacking. In these studies, we constructed Galpha-glutathione S-transferase (GST) fusion proteins and tested for interactions with [(35)S]methionine-labeled in vitro translated ZO-1 and ZO-2. Only Galpha(12) directly interacted with in vitro translated ZO-1 and ZO-2. Using a series of ZO-1 domains expressed as GST fusion proteins and in vitro translated [(35)S]methionine-labeled Galpha(12), we found that Galpha(12) and constitutively active (Q229L) alpha(12) (QLalpha(12)) bind to the Src homology 3 (SH3) domain of ZO-1. This binding was not detected with SH3 domains from other proteins. Inducible expression of wild-type alpha(12) and QLalpha(12) in Madin-Darby canine kidney (MDCK) cells was established using the Tet-Off system. In Galpha(12)-expressing cells, we found that ZO-1 and Galpha(12) co-localize by confocal microscopy and co-immunoprecipitate. Galpha(12) from MDCK cell lysates bound to the GST-ZO-1-SH3 domain, and expression of QLalpha(12) in MDCK cells reversibly increased paracellular permeability. These studies indicated that ZO-1 directly interacts with Galpha(12) and that Galpha(12) regulates barrier function of MDCK cells.

To provide a barrier function, epithelial cells have evolved a highly organized junctional complex that includes tight junctions (TJs), 1 adherens junctions, gap junctions, and desmosomes. A variety of proteins have been identified within the junctional complex; they include integral membrane proteins (such as claudins, occludin, and E-cadherin), signaling molecules (protein kinase C, Src tyrosine kinases, small G proteins, and heterotrimeric G␣ subunits), and potential scaffolding proteins (ZO-1, ZO-2, and ZO-3; for reviews, see Refs. 1 and 2). The ZO proteins are multidomain MAGUK (membrane-associated guanylate kinase) family members that contain a potential guanylate kinase domain, multiple PDZ domains, and a SH3 domain (see Fig. 1). Although ZO-1 is expressed in multiple cell types, in epithelial cells it is localized only in the tight junction, the most apical component of the junctional complex. The two related family members, ZO-2 and ZO-3, form a complex with ZO-1 in the TJ and link to the actin cytoskeleton (3). The function of ZO proteins in regulating the junction is unknown, but based upon their domain structure, they have been proposed to be scaffolding proteins (4).
Several G␣ subunits, including G␣ 12 , partially co-localize within the epithelial cell junction (1). Regulation of the junctional complex by heterotrimeric G proteins was suggested in early experiments (5). Subsequent studies have demonstrated that pertussis toxin-sensitive family members, G␣ i2 and G␣ o , localize in the tight junction of Madin-Darby canine kidney (MDCK) cells and affect both tight junction assembly and baseline properties (6,7). Recently we showed that G␣ s stimulates tight junction assembly and co-localizes with a ZO-1 complex (8). However, to date, there has been no direct demonstration of an interaction between G protein signaling molecules and TJ proteins. Utilizing in vitro binding studies and inducible expression in MDCK cells, we demonstrate that G␣ 12 directly binds to ZO-1 through the SH3 domain and that activated G␣ 12 regulates paracellular permeability.
In Vitro Protein Binding Studies-G␣ i2 -, G␣ s -, G␣ q -, G␣ 12 -, and G␣ 13 -GST fusion protein constructs were cloned as N-terminal GST-G␣ fusion proteins using standard techniques into pGEX4T (Amersham Biosciences). GST fusion proteins were expressed in Escherichia coli and purified from bacterial lysates as described previously (11). In vitro translation of the G␣ subunits, ZO-1, and ZO-2 was performed using [ 35 S]methionine in a coupled rabbit reticulocyte lysate system (Promega, Madison, WI) as described previously (11). Similar amounts of 35 S-labeled proteins were incubated with 1 g of GST fusion protein * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  prebound to glutathione-agarose 4B overnight at 4°C in 50 mM Tris-HCl, pH 7.4, 75 mM sucrose, 6 mM MgCl 2 , 1 mM EDTA, 1 mM dithiothreitol, 0.1% Triton X-100 (buffer A). Samples were washed in buffer A, eluted, and analyzed by SDS-PAGE and autoradiography.
Tet-Off Inducible MDCK Cell Lines-wt␣ 12 and QL␣ 12 were subcloned into the pTRE expression vector (CLONTECH) by standard techniques. Tet-Off MDCK cells were cultured as described by Jou et al. (12). Cells were transfected using linearized plasmid and pTK-Hyg with FuGENE 6 (Roche Molecular Biochemicals) according to the manufacturer's instructions. Resistant clones were maintained in medium containing 100 g/ml hygromycin, 100 g/ml G418, and 40 ng/ml doxycycline. Subsequent experiments were carried out in medium without hygromycin and G418 and with or without doxycycline.
Immunohistochemistry-G␣ 12 -and QL␣ 12 -transfected Tet-Off MDCK cells were grown to confluency with or without doxycycline for specific times. Cells were fixed with ice cold 100% methanol for 20 min at Ϫ20°C and blocked for 45 min at room temperature in 5% (w/v) nonfat milk containing 0.05% Triton X-100. G␣ 12 antibody was diluted 1:100 into ZO-1 hybridoma and incubated at 4°C for 1.5 h and rinsed three times with 0.05% Triton X-100 followed by incubation with goat antirabbit Texas Red and goat anti-rat FITC-conjugated secondary antibodies (Pierce) at a 1:100 dilution. Slides were viewed on a Bio-Rad MRC-1024/2p confocal microscope, and images were processed using Photoshop software (Adobe, San Jose, CA).
Immunoprecipitation and Western Immunoblot Analysis-wt␣ 12 or QL␣ 12 cells were cultured with or without doxycycline for 3 days. Whole cell lysates were obtained by scraping monolayers in buffer B (100 mM NaCl, 2 mM EDTA, 10 mM HEPES, pH 7.5, 1 mM NaVO 4 , 25 mM NaF, 1 mM phenylmethylsulfonyl fluoride, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS) and brief sonication on ice. Antibodies against ZO-1 (rat) or G␣ 12 (rabbit) were added to the lysate overnight at 4°C. Beads were washed with buffer B, and proteins were eluted in SDS-PAGE sample buffer. Western blot analysis using chemiluminescence detection (Pierce) was performed as described previously (13).
Transepithelial Resistance (TER) and Paracellular Flux Measurements-MDCK cells were plated on polycarbonate filters (Transwell, Costar) at confluent density (ϳ2 ϫ 10 5 cells/cm 2 ), and TER was measured at different time points using a Millipore ERS electrical resistance system. Measurements are expressed in ohm ϫ cm 2 as a mean of the original readings after subtraction of background values. Paracellular flux was measured as described previously (14) using cells cultured on Transwell filters for 3 days with or without doxycycline. FITC-labeled 70-kDa dextran (concentration, 3.5 M; Molecular Probes, Eugene, OR) was added into the apical chamber, and aliquots were taken from the lower chamber after 20, 40, and 60 min. FITC-dextran concentrations were determined using a fluorescent plate reader and standard curves (excitation, 485 nm; emission, 530 nm; Cytofluor 2300, Millipore Corp., Waters Chromatography, Bedford, MA) as described previously (14).

RESULTS AND DISCUSSION
The function(s) for multiple ZO proteins within the TJ is unknown, but other PDZ family members have been shown to provide a scaffold for signaling complexes. In Drosophila, a G protein signaling complex has been identified in association with a related PDZ protein (15). We and others have demonstrated localization of signaling molecules within the TJ by confocal microscopy (7,8,16,17), and other studies have shown immunoprecipitated ZO-1 and ZO-2 as a complex containing other proteins (9,18,19). To address whether heterotrimeric G␣ subunits can directly interact with ZO-1 or ZO-2, we tested whether GST fusion proteins of G␣ s , G␣ i2 , G␣ q , G␣ 12 , or G␣ 13 could interact with in vitro translated ZO-1 and ZO-2. We found that only GST-G␣ 12 consistently interacted with ZO-1 and ZO-2 in pull-down experiments (Fig. 1A). The failure to detect an interaction with the other G␣ subunits could result from conformational constraints of fusion proteins or differences in protein stability. To confirm the G␣ 12 /ZO-1 interaction and identify the ZO-1 domain interacting with G␣ 12 , we tested in vitro translated G␣ 12 and QL␣ 12 for interaction with a series of GST fusion proteins containing various regions of ZO-1 (kindly provided by M. Balda, University of London). Fig. 1B shows ZO-1 domain structure and GST fusion proteins A-F. Fig. 1C compares the binding of in vitro translated G␣ 12 , QL␣ 12 , G␣ i2 , and G␣ 13 with GST alone, A fusion protein (amino acids 295-633, containing PDZ3 and SH3 domains), and B fusion protein (PDZ3 domain only). There was no detectable interaction of the in vitro translated G␣ subunits with GST alone or the PDZ3 domain (Fig. 1C, B fusion protein). When normalized to the starting material, the A fusion protein interacted significantly better with G␣ 12 and QL␣ 12 (Ͼ10% bound, n ϭ 3) than with G␣ 13 or G␣ i2 (Ͻ1%). We next tested G␣ 12 and QL␣ 12 for interactions with C, D, E, and F fusion proteins and found that both G␣ 12 subunits interacted with C and D GST fusion proteins, but there was no significant interaction with F fusion protein or with GST alone (Fig. 1D). In these experiments with in vitro translated proteins, we did not detect interactions with the E fusion protein (SH3). However, when these GST binding experiments were repeated with lysates from G␣ 12 expressed in MDCK cells, there was a readily detectable interaction with the SH3 (E fusion protein) domain (see Fig. 2D). G␣ 12 proteins expressed in vitro may have a lower affinity for ZO-1 than proteins expressed in cells due to the absence of lipid modifications. To see whether SH3 domains from other proteins could interact with G␣ 12 , we tested GST-SH3 domains from several other proteins (kindly provided by Dr. Bruce Mayer) and found no detectable interactions with in vitro translated G␣ 12 (Fig. 1E).
G␣ 12 is difficult to study in cells due to its low levels of expression. Therefore, we established a model system in  13 , wt␣ i2 , wt␣ 12 , and QL␣ 12 as described under "Experimental Procedures." 1 l of each in vitro translated protein is shown on the right. A discrete separation of these G␣ subunits was consistently seen under these electrophoresis conditions. Exposure time was 72 h. D, GST and fusion proteins C, D, E, and F (1 g) were incubated with [ 35 S]methionine-labeled wt␣ 12 and QL␣ 12 (10 l) as described above. 1 l of wt␣ 12 is shown in the first lane. Exposure time was 72 h. E, [ 35 S]methionine-labeled wt␣ 12 was incubated with equivalent amounts of GST-ZO-1-A and GST-SH3 from Abl, Crk, n-Src, c-Src, and Grb2 and analyzed as described above. Exposure time was 72 h. IVT, in vitro translated.

ZO-1 Binds G␣ 12 24856
MDCK cells. Similar to other reports (12,20), we utilized Tet-Off inducible MDCK cell lines to express wt␣ 12 and QL␣ 12 . wt␣ 12 and QL␣ 12 proteins were expressed within 24 -48 h after the removal of doxycycline ( Fig. 2A). Protein levels remained constant for up to 10 days in the absence of doxycycline, and expression of G␣ 12 proteins could be resuppressed within 48 h with the readdition of doxycycline (not shown). G␣ 12 had previously been demonstrated to co-localize with ZO-1 in MDCK cells by confocal microscopy (16). Endogenous G␣ 12 was barely detectable by Western blot and confocal microscopy in the presence of doxycycline (Fig. 2, A and panels a and e in B), but after 48 h of G␣ 12 expression (Ϫdox) there was co-localization with ZO-1 in the TJ (Fig. 2B, panels b and d). Expression of QL␣ 12 in MDCK cells resulted in an altered cell phenotype and a complex staining pattern for QL␣ 12 and ZO-1 (Fig. 2B, panels  f and h). Some QL␣ 12 was localized intracellularly, and this could occur from overexpression and/or reduced affinity for the plasma membrane. Although the staining patterns of QL␣ 12 and ZO-1 were different, there appeared to be partial overlap in the lateral margins of the cell. Consistent with this observation, immunoprecipitation of G␣ 12 from both wild-type and QL-expressing cells co-precipitated ZO-1 (Fig. 2C). In the absence of G␣ 12 expression (ϩdox) or with no added antibody (Fig.  2C, last lane) there was no detectable ZO-1. The expression levels of ZO-1 were not affected by wt␣ 12 or QL␣ 12 expression (determined by Western blot, not shown), and the increased immunoreactivity in QL lysates was unique to this experiment. We next utilized the ZO-1-GST fusion proteins (Fig. 1B) and Western blots to determine whether wt␣ 12 or QL␣ 12 expressed in MDCK cells could bind to specific ZO-1 domains. The ZO-1 SH3 domain (E fusion protein) was capable of binding to both wt␣ 12 and QL␣ 12 from cell lysates prepared from Ϫdox cells (Fig. 2D). Identical to the in vitro results (Fig. 1), the B and F fusion proteins did not interact (not shown), and GST alone was negative (Fig. 2D). These results indicate that the SH3 domain is sufficient for interactions with G␣ 12 but do not exclude contributions from other ZO-1 domains.
The disrupted appearance of ZO-1 in the lateral membrane of QL␣ 12 -expressing MDCK cells (Fig. 2B) suggested that barrier function might be altered in these cells. To address this question, we measured TER and paracellular flux. Base-line TER was comparable in MDCK Tet-Off cells, uninduced wt␣ 12 , and QL␣ 12 MDCK cell lines (Fig. 3A, ϩdox at T ϭ 0). There was some variability in base-line TER as seen with the two sets of QL␣ 12 cells at T ϭ 0, but this was comparable to other reports (20). Expression of wt␣ 12 protein resulted in a small decrease in TER after 24 h but remained nearly identical to Tet-Off MDCK cells throughout the remainder of the experiment (Fig. 3A, open  and closed squares). Although we cannot exclude a subtle role for overexpressing wt␣ 12 on the junction, these effects did not correlate with wt␣ 12 protein expression (Ϯdox). However, inducing QL␣ 12 expression consistently caused a significant fall in TER to ϳ25 ohm ϫ cm 2 within 48 h (Fig. 3A, n ϭ 7). This fall in TER was reversible as switching to ϩdox medium resulted in TER returning to normal over the next 24 -48 h. There was a brief "overshoot" of base-line TER during TJ formation in QL␣ 12 MDCK cells similar to what is typically observed during TJ assembly (21). The decrease in TER was sustained in QL␣ 12 MDCK cells if doxycycline remained absent but could be reversed at any time by the readdition of doxycycline (not shown). We also measured paracellular flux of 70-kDa FITC-dextran in both G␣ 12 MDCK cell lines cultured Ϯ doxycycline for 3 days (Fig. 3B). Induction of QL␣ 12 expression led to a greater than 3-fold increase in flux rate (12.8 Ϯ 1.5-40.3 Ϯ 10.4 pmol/min ϫ cm 2 , n ϭ 7), while there was no significant change in wt␣ 12 MDCK cells Ϯ doxycycline (6.5 Ϯ 1.4 versus 9.4 Ϯ 2.5 pmol/ ZO-1 Binds G␣ 12 24857 min ϫ cm 2 , n ϭ 7, Fig. 2B, left panel). As an additional control, we counted trypan blue-positive cells in QL␣ 12 monolayers Ϯ doxycycline and found no significant difference. Taken together, these results indicate that expression of QL␣ 12 , but not wt␣ 12 , reversibly increases paracellular permeability, and these observations are consistent with the changes in ZO-1 staining pattern seen in QL␣ 12 -expressing cells.
In recent years, the identification of integral membrane proteins in the TJ has resulted in significant progress toward understanding the structural features of the barrier. Claudins are essential for barrier function (2) and for paracellular regulation of specific ions (22,23). Claudin family members interact with the scaffolding proteins ZO-1, ZO-2, ZO-3, and MUPP-1 (4,24), and there are multiple interactions of these scaffolding proteins with other tight junction proteins (occludin and junctional adhesion molecule, Ref. 25) and signaling molecules (26,27). Several different signaling pathways regulate the barrier of intact epithelia. Activation and overexpression of protein kinase C isoforms or monomeric G proteins results in disruption of junctions and increased leakiness in kidney epithelial cells (12,21,28,29). Furthermore, we have previously shown that activated G␣ i2 and G␣ s increase base-line TER in MDCK cells (7,8). The finding that activation of G␣ 12 lowers TER and increases paracellular permeability raises the possibility that multiple G proteins function in concert to regulate base-line TJ properties.
For the first time, we have demonstrated that ZO-1 directly interacts with a G protein ␣ subunit and propose that ZO-1 functions to scaffold G␣ 12 . ZO proteins may organize signaling complexes within a discrete membrane microdomain, the TJ. Considering the complex signaling pathways that regulate the TJ, we would suggest that there are likely to be other direct interactions between signaling proteins and ZO proteins. We show that G␣ 12 can also interact with ZO-2 (Fig. 1A), and other lower affinity interactions may not have been detected. The finding that wt␣ 12 and QL␣ 12 interact similarly with ZO-1 suggests that the interaction is independent of G␣ 12 conformation and supports the notion that ZO-1 scaffolds G␣ 12 . Although we cannot exclude the possibility that ZO-1 directly regulates G␣ 12 , we found that the GST-ZO-1-A fusion protein (Fig. 1B) has no effect on GTP␥S binding to G␣ 12 . 2 The mechanism(s) for G␣ 12 regulation of paracellular properties within the TJ and the role of the interaction with ZO-1 remain to be determined. One possibility is that through its interaction with ZO-1, G␣ 12 is placed within close proximity to other regulatory proteins that also bind to ZO-1. Our experiments were not designed to distinguish between a role for the ZO-1/G␣ 12 interaction and other possible G␣ 12 downstream signaling pathways (such as through Rho). G␣ 13 (67% amino acid identity to G␣ 12 ), did not interact with ZO-1 or ZO-2. Since G␣ 13 shares some downstream pathways with G␣ 12 , a comparison of G␣ 13 and G␣ 12 cell lines may permit us to distinguish the role of ZO-1 binding from other mechanisms regulating the junction.
Our results indicate that the SH3 domain of ZO-1 is sufficient for binding to G␣ 12 , and there may be a putative G␣ 12 binding motif contained within the ZO-1 SH3 domain. The cytoplasmic domain of E-cadherin binds to G␣ 12 and G␣ 13 , and although E-cadherin is expressed in the adherens junction, this interaction regulates association of the transcriptional activator ␤-catenin (30). Recently a cluster of charged amino acids within the C-terminal domain of E-cadherin has been shown to be required for G␣ 12 binding (amino acids 854 -864, Ref. 31). Comparison of these amino acids with the SH3 domain of ZO-1 identifies a highly conserved group of charged residues at amino acids 514 -518 (EX(D/E)K(D/E)). The SH3 domain of ZO-2 contains an identical sequence, and ZO-2 also binds to G␣ 12 (Fig. 1A). However, the SH3 domain of ZO-3 is lacking the last two charged amino acids, and we were unable to detect an interaction of in vitro translated ZO-3 with G␣ 12 (results not shown). Finally, a recent report identified an interaction between G␣ 12 and Hsp90, and Hsp90 also contains this identical cluster of charged residues (32). Other motifs have also been identified for G␣ 12 binding (33). Taken together, these findings support a scaffolding function for ZO-1 and suggest a similar role for ZO-2 and ZO-3 within the TJ. Unraveling the specific connections between ZO proteins and other signaling molecules will help provide the keys to understanding TJ regulation.