Connexin45 interacts with zonula occludens-1 and connexin43 in osteoblastic cells.

The relative expression of connexin43 and connexin45 modulates gap junctional communication and production of bone matrix proteins in osteoblastic cells. It is likely that changes in gap junction permeability are determined by the interaction between these two proteins. Cx43 interacts with ZO-1, which may be involved in trafficking of Cx43 or facilitating interactions between Cx43 and other proteins. In this study we sought to identify proteins that associate with Cx45 by coprecipitation in non-denaturing conditions. Cx45 was isolated with a 220-kDa protein that we identified as ZO-1. Under the same conditions, Cx43 also was isolated with anti-Cx45 antiserum from Cx45-transfected ROS cells (ROS/Cx45 cells). Cx43 antiserum could also coprecipitate ZO-1 in the transfected and untransfected ROS cells. Double label immunofluorescence studies showed that ZO-1, Cx43, and Cx45 colocalized at appositional membranes in ROS/Cx45 cells suggesting that all three proteins are normally associated in the cells. Additionally, we found that in vitro translated ZO-1 binds to the carboxyl-terminal of Cx45 indicating that there is a direct interaction between the carboxyl-terminal of Cx45 and ZO-1. These studies demonstrate that ZO-1 interacts with Cx45 as well as with Cx43, and suggest that the interaction of connexins with ZO-1 may play a role in regulating the composition of the gap junction and may modulate connexin-connexin interactions.

Gap junction channels permit the transport of low molecular weight substances between the cytoplasm of neighboring cells. Each gap junction channel is made of two hemichannels, which are composed of six subunits called connexins (1). At least 14 different connexin isoforms have been identified in the mouse. Different tissues express different combinations of connexins, which might allow the formation of heteromeric or heterotypic channels and facilitate the modulation of gap junction permeability in these cells (1). Alterations in gap junction permeability affect many biological phenomena including cellular differentiation and development, metabolic homeostasis, and electrotonic coupling of excitable tissue (2).
Osteoblastic cells express connexin43 (Cx43) 1 , Cx45 and Cx46 (3). Although Cx46 is retained within the exocytic pathway in these cells (4), Cx43 and Cx45 are expressed on the plasma membrane and form gap junction channels. Cx43 and Cx45 have different molecular permeabilities, and changes in the relative expression of Cx43 and Cx45 in bone cells can alter the permeability of gap junctions and the expression of bone matrix proteins (2,5). These results suggest that interactions among different gap junction proteins might alter cell-cell communication in cell networks thereby modulating the expression of bone matrix proteins and cell activities. However little is known about the processes through which different gap junction proteins interact.
One protein that may be involved in the trafficking or organization of gap junction proteins is the tight junction-associated protein ZO-1. Cx43 associates with ZO-1 (6, 7), a member of the membrane-associated guanylate kinase family of proteins, all of which contain three distinct amino acid motifs that mediate protein-protein interactions: up to three PSD95/Dlg/ ZO-1 (PDZ) domains, an Src homology 3 domain, and a guanylate kinase domain (8). Membrane-associated guanylate kinases bind to the carboxyl termini of membrane proteins and to internal domains of other membrane-associated guanylate kinases, thereby organizing these proteins into two-dimensional multi-protein complexes at cell-cell boundaries (9). Mutational analysis of the interaction between Cx43 and ZO-1 suggests that the second PDZ domain of ZO-1 interacts with the carboxyl-terminal region of Cx43 (6,7,10).
To understand the interactions between Cx43 and Cx45 in osteoblastic cells, in the current work we sought to identify proteins that interact with Cx45 and in particular wished to determine whether ZO-1 interacts with Cx45 as well as with Cx43. We found that ZO-1 could be isolated with Cx45 from ROS/Cx45 cells using a coimmunoprecipitation assay. Immunofluorescence studies showed that Cx45 colocalized with ZO-1 and Cx43 in the transfected cells. We found that in vitro translated ZO-1 bound to an oligopeptide corresponding to the final 12 amino acids in Cx45, suggesting that ZO-1 recognizes the carboxyl-terminal of Cx45.

MATERIALS AND METHODS
Reagents and Plasmid-Anti-Cx43 (11) and anti-Cx45 antiserum (5) were previously characterized. Polyclonal and monoclonal antibodies directed against ZO-1 were obtained from Zymed Laboratories Inc. (South San Francisco, CA) whereas the monoclonal antibody directed against Cx43 was purchased from Chemicon (Temecula, CA). Unless otherwise noted, reagents were purchased from Sigma. Myc-tagged ZO-1 (ZO-1myc) in pBluescript was a gift from Dr. Alan Fanning in the Dr. J. M. Anderson Laboratory (Yale University, New Haven, CT) (12).
Cell Culture and Transfection-ROS 17/2.8 cells are an osteosarcoma cell line that has been shown to express Cx43 and Cx46 but not Cx45 (13). Our experiments showed that these cells express ZO-1 and Ncadherin and suggest that they do not express ZO-2 or occludin (data not shown). ROS cells were cultured in minimum Eagle's medium containing 10% heat inactivated bovine calf serum containing 2 mM glutamine, 1 mM sodium pyruvate, 1% nonessential amino acids, 5 units/ml penicillin, and 5 g/ml streptomycin. The Cx45-transfected ROS cells had been reported previously (5,14). Stably transfected ROS cells were selected and cultured in the same culture media containing 400 g/ml G418.
Immunoprecipitation-The coimmunoprecipitation protocol was adapted from a procedure used to precipitate ZO-1-associated proteins (15,16). Cells were labeled with [ 35 S]methionine (100 Ci/ml) (Amersham Pharmacia Biotech or ICN, Costa Mesa, CA) for 2 h in methionine-depleted minimum Eagle's medium containing 10% calf serum. The media was removed, and the cells were harvested by scraping and subsequently were solubilized in mildly dissociating conditions with an IP buffer containing PBS, 1% Triton X-100, 0.5% CHAPS, 0.1% SDS, and protease inhibitors (1:100 dilution of a protease inhibitor mixture (Sigma)). The antigen-antibody complexes were collected with protein A-Sepharose. In most experiments the antigen-antibody complexes were washed with IP buffer prior to analysis by SDS-PAGE and fluorography. In some experiments the antigen-antibody complexes were collected with protein A-Sepharose and then washed in RIPA buffer, PBS containing 1% Triton X-100, 0.6% SDS, and protease inhibitors after they were collected with protein A-Sepharose. This immunoprecipitated material was then analyzed by SDS-PAGE and fluorography.
In some experiments the cells were not radioactively labeled, and the immunoprecipitated material was analyzed by immunoblotting.
Immunoblot-Proteins immunoprecipitated by anti-Cx45-antibodies were transferred to Immobilon-P membranes. The membranes were probed with 1:1000 dilution of monoclonal antibody directed against Cx43 (Chemicon) or ZO-1 (Zymed Laboratories Inc.). All blots were then probed with the appropriate peroxidase-conjugated secondary antibody (Roche Molecular Biochemicals or Jackson Immunoresearch, West Grove, PA)) and developed with the SuperSignal West Pico chemiluminescence system (Pierce).
Immunofluorescence Microscopy-ROS cells were cultured on coverslips to 60 -80% confluence and fixed in 50% methanol/50% acetone for 2 min at room temperature and permeabilized in 1% Triton X-100/PBS. The cells were blocked with PBS/1% Triton X-100/2% normal goat serum overnight and then incubated for 1 h in primary antibody. We used monoclonal anti-ZO-1 (Zymed Laboratories Inc.) at a 1:1000 dilution, monoclonal anti-Cx43 (Chemicon) at a 1:1000 dilution, or rabbit anti-Cx45 antiserum at a 1:1000 dilution (alone or in combination) as primary antibodies. The coverslips were washed and then incubated for an hour in secondary antibody. In these experiments we used a 1:2000 dilution of CY3-conjugated donkey anti-mouse IgG (Jackson Immunoresearch) and a 1:2000 dilution of Alexa-488-conjugated goat anti-rabbit antibody (Molecular Probes, Eugene, OR). The cells were viewed by epifluorescence on a Nikon Eclipse E600FN microscope using a 60ϫ (numerical aperture 1.4) or a 40ϫ (numerical aperture 1.3) objective, and the images were processed using Tillvision software (Martinsried, Germany). Confocal images were collected on a Bio-Rad MRC10 -24 microscope with LaserShop software, and the images were analyzed using Adobe Photoshop software.
ZO-1 Binding Assay-To mimic the carboxyl-terminal of Cx45, peptides consisting of amino acids 383-394 of chicken Cx45 (CSKSG-DGKNSVWI), which we call Cx45CT, were synthesized (Research Genetics, Huntsville, AL.) and conjugated to SulfoLink Gel (Pierce) through the added amino-terminal cysteine. Control peptides used in this study included the amino acids 373-382 of rat Cx43 (CSRPRPD-DLEI), which we call Cx43 CT, and amino acids 367-378 on chicken Cx45 (CSREKKSKAGSNK), which we call Cx45OC (which is 41.7% identical with amino acids 479 -490 in rat occludin), and amino acids 252-271 of rat Cx43 (CGPLSPSKDCGSPKYAYFNGC), which we call Cx43PET, were synthesized by Research Genetics and coupled to Sul-foLink Gel (Pierce) through the added amino-terminal cysteine. Peptides corresponding to amino acids 317-329 on rat Cx40 (CQPKEQPS-GASAGH), which we call Cx40I, were synthesized previously and coupled to SulfoLink Gel through the added amino-terminal cysteine. ZO-1myc was translated in a TNT T7 quick transcription system (Promega, Madison, WI) and divided into eight equal portions, which were incubated with 50 l of the peptide-coupled SulfoLink Gels, protein A-Sepharose or protein A-Sepharose and anti-ZO-1 polyclonal antibody. The gels were extensively washed with RIPA buffer and analyzed by SDS-PAGE and fluorography. We found that washing the resins with less stringent buffers, such as IP buffer, produced an unacceptably high level of nonspecific binding.

A 220-kDa Polypeptide Precipitates with Cx45 in ROS/Cx45
Cells-We identified proteins that associate with Cx45 in ROS/ Cx45 transfectants by immunoprecipitation of Cx45-containing complexes in mild detergent conditions (Fig. 1A). Detergent lysates were made from [ 35 S]methionine-labeled cells solubilized with IP buffer (PBS containing 1% Triton X-100, 0.5% CHAPS, and 0.1% SDS). Anti-Cx45 antibodies were added to the lysate, antigen-antibody complexes were collected with protein A-Sepharose, and the immunoprecipitated material was extensively washed with IP Buffer or the more stringent RIPA buffer (PBS containing 1% Triton X-100 and 0.6% SDS) (Fig.  1A) prior to analysis by SDS-PAGE and fluorography. As shown in Fig. 1A, the immunoprecipitated material derived from ROS/Cx45 cells revealed a 45-kDa band in both IP bufferwashed and RIPA buffer-washed samples. However, the IP buffer-washed samples also contained a 220-kDa band that was not detected in the RIPA buffer washed precipitates. As expected, nothing was detected in the IP or RIPA bufferwashed precipitates from the untransfected ROS cells (Fig.  1A). To confirm that this 220-kDa coprecipitated band occurred in cells expressing endogenous Cx45 as well as in the ROS/ Cx45 transfectant, we isolated Cx45 immunoprecipitates from the osteoblastic cell line UMR 106 -01, which expresses endogenous Cx45, and again detected a band that migrated at ϳ220 kDa (Fig. 1B). The molecular mass of this polypeptide is similar to that of ZO-1, a protein that is known to bind to Cx43.

ZO-1 and Cx43 Are Present in Cx45
Immunoprecipitates-We next asked whether the 220-kDa protein that coisolated with Cx45 was ZO-1. In the experiments shown in Fig.  2A, ROS cells, ROS/Cx45 cells, and UMR cells were harvested and solubilized in IP buffer, and the soluble material was immunoprecipitated with an anti-Cx45 antiserum. The antigen-antibody complexes were collected with protein A-Sepharose, and the precipitated material was washed with either IP buffer or RIPA buffer and analyzed by immunoblotting with a monoclonal antibody directed against ZO-1. ZO-1 was found in IP buffer-washed anti-Cx45 immunoprecipitates derived from ROS/Cx45 cells but not in the RIPA buffer-washed immunoprecipitates ( Fig. 2A). As expected, the anti-Cx45 antibody did not precipitate detectable ZO-1 from ROS cells.
Because Cx43 is a known binding partner of ZO-1 and our previous data suggest an interaction between Cx43 and Cx45 in cells expressing both connexins (14), we asked whether Cx43   FIG. 1. A 220-kDa protein coprecipitates with Cx45. A, ROS and ROS/Cx45 cells were labeled with [ 35 S]methionine and lysed in the IP buffer and Cx45, and associated proteins were isolated by immunoprecipitation with Cx45 antiserum. Immunoprecipitates were then washed with RIPA buffer or the less stringent IP buffer, and SDS-PAGE and fluorography were performed. B, ROS, ROS/Cx45, and UMR cells were labeled with [ 35 S]methionine and lysed in the IP buffer, and Cx45 and associated proteins were isolated by immunoprecipitation with Cx45 antiserum. Immunoprecipitates were then washed with IP buffer, and SDS-PAGE and fluorography were performed. was also precipitated by Cx45 antiserum in ROS/Cx45 cells. Proteins immunoprecipitated from transfected ROS cells were analyzed by immunoblotting with a monoclonal antibody directed against Cx43, showing that Cx43 was associated with the anti-Cx45 immunoprecipitate from the ROS/Cx45 cells (Fig. 2B). The anti-Cx45 antiserum did not immunoprecipitate Cx43 from the untransfected ROS cells. Similarly there was no Cx43 in the RIPA buffer-washed immunoprecipitates derived from either cell line.

ZO-1 Immunoprecipates with Cx43 in ROS and ROS/Cx45
Cells-To confirm the interactions identified above, we performed coimmunoprecipitations from radioactively labeled ROS and ROS/Cx45 cells using anti-Cx43 serum. The immunoprecipitated proteins were then analyzed by SDS-PAGE and fluorography (Fig. 3A). Anti-Cx43 precipitated both 45-kDa and 220-kDa radioactively labeled polypeptide proteins from both cell lines. There were more of both proteins in the ROS/ Cx45 cells than in the ROS cells. Another striking feature of these immunoprecipitates is that other than the 45-kDa and 220-kDa proteins there are no other [ 35 S]methionine-labeled proteins that copurify with Cx43 on these gels.
To confirm that the anti-Cx43 antiserum also immunoprecipitated ZO-1, anti-Cx43 immunoprecipitates that were derived from unlabeled ROS and ROS/Cx45 were subject to immunoblotting with a monoclonal antibody directed against ZO-1. ZO-1 was found in anti-Cx43 immunoprecipitates derived from each cell type (Fig. 3B). The immunoprecipitates derived from ROS/Cx45 cells had more ZO-1 than the immunoprecipitates derived from ROS cells. These data confirm that anti-Cx43 antiserum can isolate ZO-1 in the ROS cells.
Cx45 Colocalizes with ZO-1 and Cx43 in Cx45-transfected ROS Cells-Subsequently we examined the localization of Cx45, ZO-1, and Cx43 in the transfected ROS cells by confocal microscopy of immunofluorescently stained ROS/Cx45 cells. Cells were fixed, permeabilized, and stained with a monoclonal antibody directed against ZO-1 and a polyclonal antiserum against Cx45. These cells were than examined using laser scanning confocal microscopy. And as seen in Fig. 4, there was little staining for Cx45 in the ROS cells, whereas the ROS/Cx45 cells exhibited punctate and linear appositional membrane staining for Cx45. There was plentiful linear staining for ZO-1 at appositional membranes in both cell lines. Merging the Cx45 (Alexa-488) and ZO-1 (CY3) micrographs revealed areas where the two signals colocalized. While plasma membrane Cx45 was associated with ZO-1, some of the ZO-1 at the plasma membrane was not associated with Cx45. In contrast there was very little Cx45 staining seen in the ROS cells. Thus Cx45 at the cell surface colocalized with ZO-1, consistent with the data above suggesting that these proteins could associate.
We next confirmed that Cx43 also colocalized with Cx45 in these cells as was seen previously (14). The Cx45-transfected cells were simultaneously stained with an antibody directed against Cx43 and Cx45 as seen in Fig. 5. In ROS cells not expressing Cx45, Cx43 was present at cell-cell boundaries and some staining in cytoplasmic vesicles, while minimal staining was seen with the Cx45 antibody. The anti-Cx45 antiserum and the anti-Cx43 antibody stained the ROS/Cx45 cells in discrete spots along the plasma membrane, and a significant amount of staining was evident in a perinuclear region as well. The merged image shows that there is substantial colocalization of Cx43 and Cx45 in the ROS/Cx45 cells. These results are consistent with the hypothesis that transfected Cx45 associates with Cx43 in the ROS/Cx45 cells.
ZO-1 Binds to the Cx45 Carboxyl-terminal-ZO-1 binding usually occurs via interactions with the carboxyl terminus of membrane proteins. To determine whether ZO-1 binds to the carboxyl-terminal of Cx45 oligopeptides corresponding to the final 12 amino acids in Cx45 (Cx45CT), the carboxyl- terminal   FIG. 2. ZO-1 and Cx43 precipitate with Cx45. A, ROS, ROS/Cx45, and UMR cells were lysed in the IP buffer, and Cx45 and associated proteins were isolated by immunoprecipitation with Cx45 antiserum. Immunoprecipitates were then washed with a RIPA buffer or the less stringent IP buffer, run on SDS-PAGE, and transferred to Immobilon membrane. The membrane was probed with a monoclonal antibody against ZO-1 and peroxidase-conjugated secondary antibodies. B, ROS and ROS/Cx45 cells were lysed in the IP buffer, and Cx45 and associated proteins were isolated by immunoprecipitation with Cx45 antiserum. Immunoprecipitates were isolated in RIPA buffer or the less stringent IP buffer as above. The membrane was probed with a monoclonal antibody against Cx43 and peroxidase-conjugated secondary antibodies. FIG. 3. ZO-1 coprecipitates with Cx43. A, ROS and ROS/Cx45 cells were labeled with [ 35 S]methionine and lysed in the IP buffer, and Cx43 and associated proteins were isolated by immunoprecipitation with anti-Cx43 polyclonal antiserum. Immunoprecipitates were then washed with IP buffer, and SDS-PAGE and fluorography revealed proteins precipitated by anti-Cx43 antiserum from the corresponding cell lines. B, ROS and ROS/Cx45 cells were lysed in the IP buffer, and Cx43 and associated proteins were isolated by immunoprecipitation with anti-Cx43 antiserum. Immunoprecipitates were then washed with the IP buffer, separated by SDS-PAGE, and transferred to Immobilon membrane. The membrane was probed with a monoclonal antibody against ZO-1 and peroxidase-conjugated secondary antibodies.

FIG. 4. ZO-1 colocalize with Cx45 in ROS/Cx45 transfectants.
Cells were fixed, permeabilized, and stained as above. Confocal immunofluorescence microscopy shows the Cx45 and ZO-1 staining in ROS and ROS/Cx45 cells. In the right panels, the Cx45 and ZO-1 images were merged to show yellow areas that were positive for both Cx45 and ZO-1.
nine amino acids of Cx43 (Cx43CT), and internal peptides from Cx45 (Cx45OC) and Cx40 (Cx40I) were synthesized and conjugated to SulfoLink Gel through an added amino-terminal cysteine. Equal portions of in vitro translated [ 35 S]methioninelabeled ZO-1myc were incubated with the peptide-derivatized gel beads, protein A-Sepharose, or anti-ZO-1 antibody and protein-A-Sepharose, and these beads were then extensively washed with RIPA buffer. As can be seen in the resulting gel (Fig. 6), the Cx45CT peptide and the Cx43CT peptide bind significant amounts of ZO-1myc. In contrast to this, protein A-agarose, the Cx40I peptide-coupled agarose, and Cx45OC peptide-coupled agarose do not bind ZO-1myc. The multitude of bands on this gel are probably due to incomplete transcription of the relatively lengthy ZO-1myc mRNA as inclusion of the Sigma protease inhibitor mixture did not significantly alter the presence of these bands (data not shown); nonetheless they were recognized by the anti-ZO-1 antibody (data not shown), and thus they are portions of the ZO-1myc polypeptide. In a parallel experiment ZO-1myc did not bind to amino acids 252-271 of rat Cx43 that had been coupled to agarose (data not shown). All of these results suggest that ZO-1 recognizes the carboxyl-terminal of Cx45. DISCUSSION These studies demonstrate that ZO-1 interacts with Cx45, and that this Cx45/ZO-1 interaction probably requires recognition of the carboxyl terminus of Cx45. The primary evidence of this is that Cx45 and ZO-1 can be coprecipitated from lysates of ROS/Cx45 cells. Furthermore Cx45 and ZO-1 colocalize in transfected ROS cells as seen by immunofluorescence and confocal microscopy suggesting that these proteins may be in close proximity to each other at the plasma membrane. We also found that in vitro translated ZO-1 binds specifically to the 12 most carboxyl-terminal amino acids in Cx45. The interaction between the carboxyl terminus of Cx45 and ZO-1 is likely to occur via binding to a ZO-1 PDZ domain in the same manner as Cx43 carboxyl-terminal.
Work in other laboratories has shown that Cx43 binds to the second PDZ domain in ZO-1 (6,7). Both groups isolated the second ZO-1 PDZ domain in a yeast two-hybrid assay using the carboxyl-terminal of Cx43 as bait. Giepmans and Moolenar showed that the absolute carboxyl-terminal of Cx43 is needed to bind ZO-1; masking the carboxyl-terminal with a Myc epitope or removing the absolute carboxyl-terminal isoleucine abrogated this interaction (7), which is consistent with a canonical PDZ domain-carboxyl-terminal interaction. The car-boxyl-terminal sequence of Cx45 (NSVWI) is somewhat different from Cx43 (DDLEI), which raises the possibility that the Cx43 and Cx45 carboxyl termini might bind to different domains in ZO-1. Indeed the carboxyl termini of claudins (XXXYV), a known binding partner for the first PDZ domain in ZO-1, seem to be more similar to the Cx45 carboxyl-terminal (NSVWI). Additionally the presence of Cx43, ZO-1, and Cx45 in the same immunoprecipitates suggests that it is unlikely that Cx43 and Cx45 bind to the same PDZ domain.
It is still unclear what role ZO-1 plays in the life cycle of a connexin. One hypothesis is that ZO-1 makes a scaffold that temporarily secures the different connexins in gap junction plaques at the cell-cell boundary. The amino acid sequences of the carboxyl-terminal of all of connexins, with the exception of Cx32, end in a hydrophobic residue suggesting that they might bind to PDZ domains in ZO-1 (18). This might indicate that sets of connexins that bind to ZO-1, like Cx45 and Cx43, could be found in the same gap junctions along with ZO-1. In contrast Cx32 and Cx43 sort to different plasma membrane locales in thyroid epithelial cells, and only Cx43 colocalized with ZO-1 (19). It is also possible that interaction with a ZO-1 scaffold stabilizes the connexin at the gap junction. Recent data from Toyofuku et al. are consistent with this notion as Cx43 that cannot interact with ZO-1 (due to mutation or phosphorylation by c-Src) turns over much more rapidly than Cx43 that can interact with ZO-1 (10). In a scaffolding model, the different domains of ZO-1 serve as docking modules for kinases and phosphatases that interact with the different connexin polypeptides. ZO-1 could then serve a common function in the life cycle of a number of different connexins by providing a docking platform for these signaling molecules. FIG. 5. Immunolocalization of Cx45 and Cx43 in transfected ROS cells. ROS cells and transfected ROS cells were fixed, permeabilized, and stained with rabbit antibodies against Cx45 and a mouse monoclonal antibody against Cx43. Rabbit antibodies were seen with Alexa-488-conjugated donkey anti-rabbit IgG, and mouse antibody was revealed with a CY3-conjugated donkey anti-mouse IgG. In the right panels, the Cx45 and Cx43 images were merged to show yellow areas that were positive for both Cx45 and Cx43.
FIG. 6. In vitro translated ZO-1myc binds to the to the carboxyl-terminal of Cx45. ZO-1myc was in vitro translated in the presence of [ 35 S]methionine and then incubated with protein A-Sepharose, with protein A-Sepharose and anti-ZO-1 antibody, or agarose conjugated with Cx43CT, Cx40I, Cx45OC, or Cx45CT peptides. The beads were collected by centrifugation and extensively washed with RIPA buffer, and the bound peptides were analyzed by SDS-PAGE and fluorography.