Regulation of Connexin32 by ephrin receptors and T-cell protein-tyrosine phosphatase

Gap junctions are intercellular conduits that permit the passage of ions, small metabolites, and signaling molecules between cells. Connexin32 (Cx32) is a major gap junction protein in the liver and brain. Phosphorylation is integral to regulating connexin assembly, degradation, and electrical and metabolic coupling, as well as to interactions with molecular partners. Cx32 contains two intracellular tyrosine residues, and tyrosine phosphorylation of Cx32 has been detected after activation of the epidermal growth factor receptor; however, the specific tyrosine residue and the functional implication of this phosphorylation remain unknown. To address the limited available information on Cx32 regulation by tyrosine kinases, here we used the Cx32 C-terminal (CT) domain in an in vitro kinase-screening assay, which identified ephrin (Eph) receptor family members as tyrosine kinases that phosphorylate Cx32. We found that EphB1 and EphA1 phosphorylate the Cx32CT domain residue Tyr243. Unlike for Cx43, the tyrosine phosphorylation of the Cx32CT increased gap junction intercellular communication. We also demonstrated that T-cell protein-tyrosine phosphatase dephosphorylates pTyr243. The data presented above along with additional examples throughout the literature of gap junction regulation by kinases, indicate that one cannot extrapolate the effect of a kinase on one connexin to another.

developing spontaneous liver tumors and are more vulnerable to chemical-induced liver tumorigenesis (2,3); these mice exhibit other pathological phenotypes, such as decreased nerve-dependent bile secretion, enhanced pancreatic amylase secretion, and peripheral nerve demyelination (4 -7). Since identifying Cx32 as an X-linked Charcot-Marie-Tooth disease-causing protein, more than 400 different mutations affecting every region of the Cx32 protein have been reported (8). These mutations alter trafficking to the plasma membrane, channel gating, channel selectivity, protein partner interactions, and the phosphorylation status (9).
All stages of the connexin life cycle, which include assembly, degradation, the regulation of electrical and metabolic coupling, and regulation of interactions with molecular partners, involve phosphorylation (10). Cx32 contains 12 serine residues within the cytoplasmic domains that are enzyme-accessible (N-terminal (NT) domain, 2 residues; cytoplasmic loop, one residue; C-terminal (CT) domain, 9 residues) and is in vitrophosphorylated by Ca 2ϩ /calmodulin-dependent protein kinase II, protein kinase A (PKA), and protein kinase C (PKC) (11)(12)(13). For the MAP kinase p38, a correlation was made between the differentiation of rat mammary epithelial cells and Cx32 phosphorylation; unfortunately, direct phosphorylation of Cx32 was not tested (14). Cx32 also contains two intracellular tyrosine residues (NT, Tyr 7 ; CT, Tyr 243 ). Of note, Tyr 211 may or may not be accessible to the cytosol based on its juxtaposed location to the membrane; however, accessibility would not be favorable to protein interactions. Tyrosine phosphorylation of Cx32 was induced in purified liver membranes by the epidermal growth factor receptor (EGFR) kinase, and the presence of the Cx32binding partner calmodulin (CaM) inhibited tyrosine phosphorylation (11,15). Unfortunately, the site(s) of Cx32 phosphorylation by EGFR and its relevance to Cx32 function was not established (11). Finally, Y7 phosphorylation was observed by MS from Cx32-transfected HeLa cells; however, the kinase involved was not identified (16). Based upon the limited available information about Cx32 regulation by tyrosine kinases, we used an in vitro kinase-screening assay, which identified the ephrin type-B receptor 1 (EphB1) and ephrin type-A receptor 1 (EphA1) as novel tyrosine kinases that phosphorylate Cx32.
Eph receptors and ligands are both membrane-bound; thus, binding and activation requires cell-to-cell contact. Downstream signaling is important for proper cell sorting during development, cell adhesion, migration, repair after nervous system injury, and maintenance of gap junctions (17)(18)(19)(20). The EphB4 receptor coimmunoprecipitated with Cx43, and its activation in primary cultures of rodent cardiomyocytes inhibited gap junction intercellular communication (GJIC) (21). Another study showed that GJIC is inhibited at ectopic ephrin boundaries and that ephrin-B1 physically interacts with Cx43 and influences its distribution (19). Altogether, these studies suggest a mechanism by which Eph receptors and ligands mediate control of cell-to-cell communication through phosphorylation of the gap junction proteins. Nevertheless, whether the Eph receptors directly phosphorylate connexins and whether this is a general mechanism to regulate other connexin isoforms remain to be determined.
Here, we identified that the Eph receptor isoforms EphB1 and EphA1 phosphorylate Cx32CT residue Tyr 243 , an event that increases GJIC. We also demonstrate that the T-cell protein-tyrosine phosphatase (TC-PTP) dephosphorylates Cx32CT residue pTyr 243 . This study provides evidence that characterization of the same kinase for different connexin isoforms is important, as one cannot extrapolate the effect of a kinase on one connexin to another.

EphB1 directly interacts and phosphorylates the Cx32CT domain
GPS 2.0, NetPhos 2.0 server, and KinasePhos 2.0 were used to predict tyrosine kinases that could phosphorylate the Cx32 NT and CT domains. Five tyrosine kinases were selected for an in vitro tyrosine phosphorylation screening assay performed by Eurofins Scientific (KinaseProfiler) using the Cx32NT (Met 1 -Gly 21 ; Tyr 7 ) or the Cx32CT (Cys 217 -Cys 283 ; pTyr 243 ) as substrates (Fig. 1A). Of the kinases tested, EphB1 and Ron resulted in phosphorylation of the Cx32CT greater than 50 and 40% of control peptide signal, respectively. EphB1 was the only kinase able to phosphorylate the Cx32NT and did so to a level similar to that of the Cx32CT. To validate the screen results, the kinase assay was repeated using EphB1, Ron, and EGFR in our laboratory. Consistent with data from Eurofins Scientific, EphB1 and Ron phosphorylated Cx32CT residue pTyr 243 but not EGFR (Fig. 1B). Of note, work from Diez et al. (11) demonstrated that EGFR can phosphorylate immunoprecipitated Cx32 in vitro as detected by autoradiography.
To confirm the Cx32 interaction with the EphB1 and Ron kinases, a pulldown assay was performed using purified GST, GST-Cx32NT, or GST-Cx32CT and lysate from HeLa cells expressing EphB1 and Ron (but not Cx32) (Fig. 1C). EphB1 was pulled down by GST-Cx32CT, but not GST or GST-Cx32NT. The EphB1 interaction with the Cx32CT was stabilized by the chemical cross-linker 3,3Ј-dithiobis(sulfosuccinimidylpropionate) (DTSSP), which can capture weak and/or transient interactions (22)). Co-immunoprecipitation (co-IP) showed that EphB1 is in the same complex with full-length Cx32 (Fig. 1D). Conversely, whereas Ron could phosphorylate the Cx32CT in vitro, the interaction was weak in the pulldown assay (even in the presence of DTSSP) and not detected by co-IP (data not shown). Thus, Ron was excluded from further experimentation, highlighting a caveat in this type of in vitro screen where the phosphorylation in vitro does not always correlate with a detectable interaction in cells.

TC-PTP interacts with and dephosphorylates Cx32CT residue pTyr 243
Previously, we identified that TC-PTP directly dephosphorylated Cx43 on the CT domain, leading to increased GJIC (23). A, the Cx32NT and CT domains were used as a substrate for the human EGFR, Blk, EphB1, Met, and Ron kinases (performed by Eurofin Scientific). The amount of Cx32NT and CT phosphorylation was compared with a positive control peptide for each kinase (percentage of control signal). The dashed line highlights 50% of the control signal. B, an in vitro kinase assay for the Cx32CT was performed in our laboratory to repeat the kinase screen performed by Eurofin Scientific (EphB1, Ron, and EGFR). A general anti-phosphotyrosine antibody was used to detect the phosphorylation level by Western blotting (WB). The control group used buffer to substitute the kinases in the reactions. C, purified GST alone, GST-Cx32CT (CT), or GST-Cx32NT (NT) bound on GSH-agarose beads was incubated without (Ϫ) or with (ϩ) HeLa cell lysate, and the pulled down proteins were analyzed by Western blotting using anti-EphB1 or anti-Ron antibodies. The chemical cross-linker DTSSP was used to capture the weak and/or transient interactions. D, co-IP was performed in HeLa cells stably expressing Cx32. Lysates were immunoprecipitated with anti-Cx32 or IgG and then blotted for Cx32 and EphB1. Error bars, S.D.

Cx32 (de)phosphorylation
Whether TC-PTP can dephosphorylate other connexin isoforms (e.g. Cx32 pTyr 243 ) is unknown. To determine whether a direct interaction exists between TC-PTP and the Cx32CT domain, NMR titration experiments were performed with the purified TC-PTP catalytic domain (TC-PTP(1-314)) and Cx32CT. Different concentrations of unlabeled TC-PTP(1-314) were titrated into 15 N-labeled Cx32CT (residues Cys 217 -Cys 283 ) and 15 N HSQC spectra were acquired ( Fig. 2A). By the 1:1 molar ratio, Cx32CT residues Cys 217 -Ser 266 (including Tyr 243 ) had broadened beyond detection (highlighted on the Cx32CT sequence in Fig. 2B; red). The decrease in signal intensity from increasing TC-PTP(1-314) concentrations was fit according to the nonlinear least square method, and the binding affinity (K D ) for the interaction was determined to be 19.9 Ϯ 4.5 M (Fig. 2C). A GST pulldown experiment was performed to validate the NMR results. Purified GST alone, GST-Cx32CT, or GST-Cx32NT bound to GSH-Sepharose was incubated with purified TC-PTP(1-314). Immunoblotting using an anti-TC-PTP NT antibody identified that GST-Cx32CT, but not NT, directly interacted with the TC-PTP catalytic domain (Fig. 3, A  and B). GST-Cx32CT also pulled down full-length TC-PTP from HeLa cell lysate (Fig. 3C).
To determine whether TC-PTP dephosphorylates the Cx32 on residue Tyr 243 , an in vitro phosphatase assay was conducted using a peptide (His 237 -Asn 249 ) containing pTyr 243 incubated with TC-PTP(1-314). Following the protocol for the Malachite green assay (Millipore), an increase in the amount of P i production was observed, indicating that TC-PTP dephosphorylated the Cx32 peptide on pTyr 243 (Fig. 3D). Notably, TC-PTP has a comparable affinity for both the Cx32 peptide and positive control, as shown by the kinetic data in Table 1

EphB1 phosphorylates and TC-PTP dephosphorylates Cx32 in HeLa cells
We have shown that EphB1 phosphorylates and TC-PTP dephosphorylates Cx32 in vitro. To determine whether EphB1 and TC-PTP affect the level of Cx32 pTyr 243 in cells, HeLa cells stably expressing Cx32 WT were transfected with EphB1 and/or the TC-PTP catalytic domain (residues 1-314), and the tyrosine phosphorylation level was detected on immunoprecipitated Cx32 (Fig. 4A). No tyrosine phosphorylation of Cx32 was detected in the control lane (no transfection of EphB1 or TC-PTP(1-314)). Expression of EphB1 increased Cx32 tyrosine phosphorylation and double transfection with TC-PTP(1-314) decreased Cx32 tyrosine phosphorylation when compared with only EphB1.

Cx32 (de)phosphorylation
Although an interaction between Cx32NT and EphB1 was not observed by GST pulldown (Fig. 2), phosphorylation of Cx32NT residue Tyr 7 by EphB1 was observed in vitro; thus, we tested whether this residue could be phosphorylated in HeLa cells (16). Cx32 was immunoprecipitated from HeLa cells stably expressing Cx32 WT or Y243F, and the level of tyrosine phosphorylation was measured (Fig. 4B). Expression of EphB1 increased Cx32 WT phosphorylation, which was further increased when EphB1 was activated by the addition of the ligand, EphB2-Fc. However, this effect was not observed in cells expressing the Cx32 Y243F mutant, suggesting that EphB1 only phosphorylates Cx32CT on residue Tyr 243 (not Tyr 7 or Tyr 211 ).

Cx32 Y243E increases GJIC
Well-appreciated in the gap junction field is the use of Asp or Glu substitutions as a mimetic for phosphorylation (e.g. see Refs. 24 and 25). These substitutions allow for the precise control of the site(s) modified and the production of enough phosphorylated protein to observe the intended biological response. Here, HeLa cells either transiently or stably expressing Cx32 WT or Y243E were used to determine the functional significance of Tyr 243 phosphorylation on GJIC. Western blotting determined that the expression level of Cx32 WT was twice that of Y243E (Fig. 5A). Next, immunofluorescence was used to identify the cellular location of Cx32 (Fig. 5B). The Cx32 Y243E

Cx32 (de)phosphorylation
mutant formed gap junctions similar to WT at the plasma membrane (white arrows). To determine the effect on cell-tocell communication, the junctional transfer of the fluorescent tracer Lucifer yellow (M r 443; Ϫ2 charge) was measured in a scrape-loading assay (Fig. 5C). Interestingly, whereas Cx32 WT protein expression was double that of Y243E and formed plaques of similar size to Y243E, the level of intercellular communication is significantly increased for Y243E. The data suggest that the phosphorylation state of Tyr 243 can modulate GJIC.

EphA1 also phosphorylates Cx32 on the CT domain
Having established that EphB1 phosphorylates Cx32, we next sought to determine whether different Eph receptor family members could also phosphorylate Cx32. To address this, we again used purified recombinant Cx32CT (Cys 217 -Cys 283 ; Tyr 243 ) as a substrate for an in vitro kinase screen to other Eph receptor family members (Eurofins Scientific; KinaseProfiler). EphB1 and Yes were used as positive and negative controls, respectively. Interestingly, the only other Eph receptor to phosphorylate the Cx32CT was EphA1, which yielded ϳ67% of control peptide signal (Fig. 6A). HeLa cells stably expressing Cx32 were then transiently transfected with cDNA coding human EphA1, and after expression, the receptor was activated with soluble ephrin-A1-Fc. Similar to EphB1, activation of EphA1 caused a significant increase in Cx32 tyrosine phosphorylation (Fig. 6B). Next, we determined whether phosphorylation of Cx32 by EphA1 would occur in Caco2 cells (human colon cancer), which endogenously expresses both proteins (Fig. 6C). Similar to when overexpressed in HeLa cells, a significant increase in Cx32 tyrosine phosphorylation was observed when the Caco2 cells were stimulated with the soluble ephrin-A1 ligand (Fig. 6D). The Cx32 Y243E dye transfer data from Fig. 5 would suggest that phosphorylation of Cx32 by EphA1, like EphB1, would lead to increased GJIC. Indeed, when we tested this using Caco2 cells stimulated with the soluble ephrin-A1 ligand, a significant increase in dye transfer was observed when compared with nonstimulated cells (29 versus 12 cells receiving dye; Fig. 6E). This increase was similar to the Cx32 Y243E cells.

Phosphorylation at Tyr 243 has little to no effect on the Cx32CT interaction with CaM or synapse-associated protein 97 (SAP97)
A clear relationship exists between phosphorylation of connexin CT residues and the binding of molecular partners, with examples of both increasing and decreasing binding affinities (e.g. see Refs. 26 and 27). CaM and SAP97 were previously shown by our laboratory to interact with the Cx32CT in a region covering Tyr 243 (CaM: Cys 217 -Ser 266 ; SAP97: Cys 217 -Gln 245 ) (15). Therefore, we tested using NMR 15 N HSQC titrations whether phosphorylation of Tyr 243 would affect these interactions. The data indicate that phosphorylation of Tyr 243 alone does not have a significant impact on the Cx32CT interaction with CaM or SAP97 (Figs. S1 and S2).
The possibility exists that other kinases activated downstream of the Eph receptors would phosphorylate Cx32CT to affect binding of these proteins (e.g. all in vitro: Ser 229 (PKC), Ser 233 (PKA and PKC), and Ser 240 (PKC)) (12,16). Another possibility is that the main regulatory role of Tyr 243 phosphorylation is to affect channel properties. It is tempting to speculate that a mechanism by which CaM mediates the closure of Cx32 gap junction channels is by masking the effect of the phosphate on Tyr 243 to increase GJIC (28, 29) or preventing phosphorylation in the first place, as was observed in a study showing that tyrosine phosphorylation of Cx32 is inhibited in the presence of CaM (30).

Discussion
A few studies have investigated the regulation of GJIC by Eph receptors and ephrins; although none have looked at the Cx32 isoform, interestingly, their observations of blocking communication through gap junctions were all the same. For example, the EphB4 receptor co-immunoprecipitated with Cx43, and its activation in primary cultures of rodent cardiomyocytes inhibited GJIC (21). Another study found that ephrin-B1 co-immunoprecipitated with Cx43, and the interaction between EphB2and ephrin-B1-expressing NIH3T3 cells inhibited GJIC and redistributed Cx43 from the plasma membrane to the cytoplasm (19). Consistent with these observations is that expression of Eph receptors and ephrins in zebrafish cap cells blocked GJIC at the boundary between both cell populations (20). Finally, although GJIC was not evaluated, the presence of Cx36

Cx32 (de)phosphorylation
was necessary for the effects of EphA on insulin secretion by pancreatic cells (31).
In this study, we identified that EphB1 and EphA1 phosphorylate Cx32CT residue Tyr 243 . Phosphomimetic Cx32 (Y243E) enhanced cell-cell communication in HeLa cells without noticeably affecting the number and size of the gap junction plaque. This observation was substantiated in Caco2 cells, which endogenously express EphA1 (activated by ephrin-A1-Fc) and Cx32. Of note, Cx32 residue Ser 233 is also associated with increased GJIC when phosphorylated (12). This finding is in contrast to tyrosine phosphorylation of the Cx43CT, as phosphorylation of Tyr 247 , Tyr 265 , and Tyr 313 is associated with decreased GJIC (10,(32)(33)(34)(35)(36). Taken together, this information suggests that phosphorylation (even by the same kinase) regulates different connexin isoforms via different mechanisms. Growing evidence to support this is apparent in the case of regulation by serine kinases, as little correlation exists between Cx43 and Cx45. For example, activation of PKA decreases Cx45 and increases Cx43 junctional conductance (37,38), and activation of cGMP-dependent protein kinase has no effect on Cx45 but decreases Cx43 junctional conductance (37,38). Additionally, the response of Cx37 channels to PKC activation or inhibition is opposite to that observed for Cx43 gap junction channels (39).
An important question arises: Where in human anatomy would phosphorylation of Cx32 by Eph receptors occur? Based upon data provided by the Human Protein Atlas, Cx32 is expressed in the appendix, liver, gallbladder, gastrointestinal tract, kidney, and male/female tissues (40) (www.proteinatlas. org/) 4 (51). All of these tissues also express the EphA1 receptor at a similar level. Conversely, the overall expression of EphB1 is relatively low throughout the body; however, like Cx32, EphB1 is highly expressed in the male testis. Activation of EphA1 and EphB1 receptors by their respective ephrins (41) is typically mediated by contact-dependent communication between cells. This is similar to the cell-to-cell contact required for the formation of gap junction channels. Thus, we would hypothesize that under such conditions, phosphorylation of Cx32 would increase GJIC. Additionally, we would expect in any pathological condition where either expression of these proteins is up-regulated or cells that do not normally interact with each other do, Cx32/Eph receptor signaling could occur (18,42,43). Although future studies are necessary to address the exact context of Cx32/Eph receptor signaling in human physiology, this study does demonstrate that phosphorylation of the Cx32 residue Tyr 243 , regardless of the kinase, would increase GJIC.
The use of MS, phospho-specific antibodies, mutant versions of connexins eliminating/mimicking phosphorylation, and identification of CT phosphorylation sites has led to the detec-

Figure 6. EphA1 phosphorylation of Cx32 increases gap junction intercellular communication in vitro and in cyto.
A, Cx32CT was used as a substrate for a number of different Eph receptor isoforms. EphB1 and Yes were used as positive and negative control, respectively. The amount of Cx32CT phosphorylation was compared with a positive control peptide for each kinase (percentage of control signal). The dashed line highlights 50% of the control signal. B, HeLa cells stably expressing Cx32 were transfected with EphA1. Ephrin-A1-Fc was used to activate EphA1 receptor. Cx32 was immunoprecipitated, and the tyrosine phosphorylation level was detected by Western blotting. **, p Ͻ 0.01 (one-way ANOVA); ##, p Ͻ 0.001 compared with EphA1. Black arrows, IgG light chain. C, Western blot analysis of endogenous expression levels of EphA1, EphB1, and Cx32 in Caco2 cells. D, Cx32 was immunoprecipitated from Caco2 cells stimulated with ephrin-A1-Fc and probed for tyrosine phosphorylation by Western blotting. E, Caco2 cells cultured on glass coverslips were stimulated with ephrin-A1-Fc and scrape-loaded with Lucifer yellow (LY, green) and Texas Red-conjugated dextran (red). Nuclei were stained by DAPI to show the confluence of cells. Scale bar, 100 m. ***, p Ͻ 0.0001 (t test). Error bars, S.D.

Cx32 (de)phosphorylation
tion of phosphorylation-dependent changes in connexin structure, localization, and interacting partners as well as gap junction channel assembly, disassembly, and gating properties (for a review, see Refs. 10 and 44). Through identifying the various cellular conditions and signaling pathways involved in connexin phosphorylation, an understanding of the sequential activation of multiple kinases and changes in the pattern of connexin phosphorylation over time has emerged (e.g. see Ref. 10). The significance of defining a connexin phosphorylation pattern under different homeostatic and pathological conditions has the potential to translate this into a therapeutic intervention. Clearly, our understanding of connexin regulation will not be complete until all kinases involved and sites phosphorylated have been identified. The importance of this study is that it directly links phosphorylation of a specific Cx32 tyrosine residue to intercellular communication and provides support for the idea that characterization of the same kinase for different connexin isoforms is important, as the functional significance cannot be inferred from another isoform.

Cell culture and treatments
Cx32 WT, Y243F, and Y243E were stably expressed in HeLa cells (ATCC). HeLa and Caco2 cells were grown in Dulbecco's modified Eagle's medium (Hyclone, Thermo Fisher Scientific Inc.) supplemented with 10% fetal bovine serum (Hyclone, Thermo Fisher Scientific Inc.) and antibiotics. FHC cells were cultured in Dulbecco's modified Eagle's medium/F-12 supplemented with 25 mM HEPES, 10 ng/ml cholera toxin, 0.005 mg/ml insulin and transferrin, 100 ng/ml hydrocortisone, 20 ng/ml human recombinant EGF, 10% fetal bovine serum, and antibiotics. All cells were cultured in an atmosphere of humidified 5% CO 2 at 37°C. Where appropriate, cells were transiently transfected with either EphA1, EphB1, or a combination of either in conjunction with TC-PTP(1-314) using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. For single and combination transfections, the amount of plasmid for the Eph receptor and TC-PTP(1-314) was equivalent. Cells were treated for 1 h at 37°C with 1 g/ml either EphA1-Fc or EphB2-Fc (SinoBiological) ϳ24 h after transfection.

Antibodies and immunostaining
The following antibodies were used in this study: Cx32 mAb (Sigma-Aldrich); Cx32 polyclonal antibody (Life Technologies); EphB1 and EphA1 antibodies (Cell Signaling); TC-PTP antibodies against the N terminus and C terminus (Sigma-Aldrich); nonspecific phosphotyrosine antibody (Cell Signaling). Cells were immunostained as described previously (45). Briefly, cells grown on coverslips to ϳ60% confluence were fixed with 2% paraformaldehyde for 15 min. Cells were blocked for 30 min at room temperature (RT) by MPS buffer (1ϫ PBS, 1% goat serum) containing 0.2% TX-100 for permeabilization. Then cells were immunostained with appropriate primary antibodies at RT for 1 h, followed by several PBS plus 0.5% Tween washes. Secondary antibodies (Alexa 594 -conjugated goat anti-rabbit antibody and/or Alexa 488 -conjugated goat antimouse antibody) were applied for 1 h at RT. Images of immu-nostained cells were acquired with a Zeiss 510 Meta confocal laser scanning microscope using a ϫ63, 1.4 numerical aperture objective with appropriate filters.

Co-IP and Western blotting
Cx32-expressing HeLa cells were lysed in complete lysis buffer (50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 0.5% sodium deoxycholate, 1% TX-100, 5 mM NaF, and one-half tablet of Complete protease inhibitor (Roche Applied Science) in 20 ml of buffer), maintained on ice for 30 min, precleared with protein G beads for 30 min at 4°C, and then spun at 12,000 rpm for 15 min. Total protein was assessed using the BCA protein assay kit (Pierce). 2 mg of protein lysate was incubated with 2 g of anti-Cx32 or mouse IgG (4 h at 4°C) and then incubated with 100 l of protein G-Sepharose (GE Healthcare) (overnight at 4°C). The Sepharose was washed four times with cold lysis buffer, and the co-IP was analyzed by SDS-PAGE and Western blotting. Anti-EphB1 mAb was used to detect EphB1 co-IP with Cx32.
Protein levels were detected by SDS-PAGE and Western blotting. One exception for detecting tyrosine phosphorylation was that 5 mM Na 3 VO 4 was added in the blocking buffer and primer antibody buffer to minimize the loss of phosphorylation. Western blotting data were scanned and quantified using ImageJ as described (45).

GST pulldowns
The GST pulldown assay was modified from that described previously (46). Briefly, purified GST, GST-Cx32NT, or GST-Cx32CT was bound to GSH-Sepharose beads in buffer containing 25 mM Tris-HCl (pH 7.4),150 mM NaCl, 0.1% TX-100, 1 mM DTT, 0.1 mM EDTA, and one-half tablet of Complete protease inhibitor in 20 ml of buffer. 5 mg of either purified TC-PTP(1-314) was incubated with the GST control or either GST fusion protein overnight at 4°C on a rotating wheel. Beads were washed five times with buffer, and bound proteins were eluted with SDS-PAGE sample buffer and analyzed by Western blotting.
To pull down Cx32-binding partners, 4 mg of MDA-MB-231 cell lysate protein (in cell lysis buffer containing 25 mM Tris-HCl (pH 7.4), 150 mM NaCl, 0.5% TX-100, 0.5% sodium deoxycholate, 2 mM EDTA, PhosSTOP (Roche Applied Science), and Complete protease inhibitor) were incubated with GST-Cx32CT or the GST control for 6 h at 4°C on a rotating wheel. Beads were washed five times with cell lysis buffer, and bound proteins were eluted with SDS-PAGE sample buffer and analyzed by Western blotting.

NMR-HSQC experiment
NMR data were acquired using a 600-MHz Varian INOVA NMR spectrometer outfitted with a Bruker cryoprobe at the NMR Facility of the University of Nebraska Medical Center. Cx32CT(217-283) and TC-PTP(1-314) were expressed and purified as described previously (15,23). Gradient-enhanced two-dimensional 15

In vitro phosphatase assay
Cx32 phosphopeptide pTyr 243 (HRLSPEpYKQNEIN) was used in the Malachite green assay (Millipore) to detect whether they are substrates of TC-PTP in vitro. Positive and negative controls were DADEpYL (described previously (49)) and ARKRIpYPP (described previously (50)), respectively. Peptides and enzyme were dissolved and diluted in reaction buffer (60 mM HEPES (pH 7.2), 150 mM NaCl, 1 mM EDTA, 0.17 mM DTT, 0.83% glycerol, 0.017% BSA, and 0.002% Brij-35) and temperature-equilibrated for 15 min at RT. A 500 M concentration of each peptide was added into a 96-well plate, and then TC-PTP(1-314) was added to each peptide at different time points. The final concentration of TC-PTP(1-314) was 0.6 nM. All reactions were terminated by adding 100 l of Malachite green solution and incubated for 15 min at RT to allow color development. Absorbance at 650 nm was read using a Spectra-Max 190 spectrometer (Molecular Devices).
The Malachite green assay was also used to measure the kinetic parameters of dephosphorylation (49). The reaction was performed in 96-well plates with a final volume of 25 l. A standard curve of KH 2 PO 4 , used for calculating the release of P i , was determined on the same plate as the reaction samples. Different concentrations of peptides, [substrate], were combined with 1.2 nmol of TC-PTP(1-314). Time points of the reaction were sampled from 0 to 15 min to calculate the initial velocity (v) versus [substrate]. A Lineweaver-Burk double-reciprocal plot (rearranged from the Michaelis-Menten equation) was created based on 1/v and 1/[substrate]. The kinetic parameters were determined by the linear equation from the Lineweaver-Burk double-reciprocal plot. All of the reactions were conducted at pH 7.5, 25°C. The reported results were calculated from three independent experiments.

Scrape-loading assay
Cells were scrape-loaded as described (15). Briefly, both HeLa cells expressing Cx32 WT or the Y243E mutant and Caco2 cells were seeded on coverslips 24 h before performing the assay. Cell culture medium from 100% confluent cells was removed and replaced with 1 ml of PBS containing 0.25% Lucifer yellow and Texas Red-conjugated fluorescent dextran (10 kDa, 1.5 mg/ml; fixable). Cells were scrape-loaded with a sterile scalpel by two longitudinal scratches and then incubated at RT for 1 min for HeLa cells and 4 min for Caco2. Cells were washed quickly three times with warm PBS (containing MgCl 2 and CaCl 2 ) or cell culture medium followed by incubation at 37°C for 5 min. After incubation, cells were washed two times with warm PBS and fixed with 3.7% buffered paraformaldehyde for 15 min. 0.1 M glycine was used to quench autofluorescence for 15 min. Coverslips were mounted on glass slides in a droplet of SlowFade (Invitrogen). The result was confirmed by repeating the experiment three times and for each trial, capturing four side-by-side images. The method to quantify dye transfer was described previously (15). Briefly, cells with Lucifer yellow were counted, whereas the cells with dextran, which indicated the initially loaded cells, were excluded.

Statistical analysis
All data were analyzed by GraphPad Prism version 5.0 and presented as the mean Ϯ S.D. Either Student's t test or one-way ANOVA with Neuman-Keuls post hoc analysis was used to compare differences between all groups where appropriate. p Ͻ 0.05 was considered to be statistically significant.