Ubiquitin protein ligase Nedd4 binds to connexin43 by a phosphorylation-modulated process.

Connexin43 is degraded by the proteasomal as well as the lysosomal pathway with ubiquitin playing a role in both degradation pathways. So far, no ubiquitin protein ligase has been identified for any of the connexins. By using pull-down assays, here we show binding of a ubiquitin protein ligase, Nedd4, to the C-terminus of connexin43. This observation was confirmed in vivo by coimmunoprecipitation and immunofluorescence, showing colocalization of Nedd4 and connexin43. Binding of Nedd4 to its interaction partners is generally carried out by its WW domains. Our results indicate that the interaction with connexin43 occurs through all three WW domains of Nedd4. Furthermore, whereas WW1 and WW2 domains mainly interact with the unphosphorylated form of connexin43, WW3 binds phosphorylated and unphosphorylated forms equally. In addition, using the surface plasmon resonance approach we show that only the WW2 domain binds to the PY motif located at the C-terminus of connexin43. Suppression of Nedd4 expression with siRNA resulted in an accumulation of gap junction plaques at the plasma membrane, suggesting an involvement of the ubiquitin protein ligase Nedd4 in gap junction internalization.


Introduction
Gap junctions are accumulations of intercellular channels embedded in plasma membranes. These channels enable water, ions, second messengers and molecules of less than ~1 kDa to diffuse between adjacent cells (Goodenough et al., 1996;Kumar and Gilula, 1996). Gap junction biogenesis requires the oligomerization of six connexin proteins into a hexameric connexon and their subsequent trafficking to the plasma membrane (Beyer et al., 1990;Bruzzone et al., 1996). End-toend interactions between two connexons of neighbouring cells result in complete junctional channels. To date, 21 connexin genes have been identified in humans (Sohl and Willecke, 2004), connexin43 (Cx43) being the most widely expressed and best-studied gap junction protein. All connexins contain four transmembrane domains and their N-and C-termini are located intracellularly (Goodenough et al., 1996;Yeager et al., 1998;Evans et al., 1999). Although the two extracellular loops are structurally conserved, the cytoplasmic loop and the Cterminal tail are highly divergent. The C-terminal tail contains multiple serine and tyrosine phosphorylation sites. Several serine/threonine and tyrosine protein kinases such as protein kinase C and Src, directly phosphorylate Cx43 at particular residues, and phosphorylation generally correlates with a disruption of the intercellular junction communication (Lampe et al., 2000;Lin et al., 2001). Furthermore, epidermal growth factor (EGF) markedly increases the phosphorylation of Cx43 on Ser255, Ser279 and Ser282 (Warn-Cramer et al., 1996;Warn-Cramer et al., 1998). This phosphorylation is directly due to mitogen-activated protein kinases (MAPK) and results in a rapid and transient disruption of gap-junctional communication (Lau et al., 1992;Kanemitsu and Lau, 1993;Cottrell et al., 2003). In addition, EGF-induced phosphorylation of Cx43 increases its ubiquitylation at the plasma membrane and induces proteasome-dependent degradation of Cx43 gap junctions (Leithe and Rivedal, 2004a;Leithe and Rivedal, 2004b). Indeed, phosphorylation of connexins seems to play an important role in regulating the gating, the assembly and the degradation of gap junction channels.
Little information is so far available on the processes responsible for degradation of integral membrane proteins. It has been demonstrated that several different ion channel proteins interact through binding of their C-terminal PY motif (xPPxY) to WW domains of Nedd4 ubiquitin protein ligases, leading to their ubiquitin-dependent targeting for degradation (Staub et al., 1996;Staub et al., 1997;Abriel et al., 2000;Schwake et al., 2001;Ingham et al., 2004). WW domains are Connexin43 is degraded by the proteasomal as well as the lysosomal pathway with ubiquitin playing a role in both degradation pathways. So far, no ubiquitin protein ligase has been identified for any of the connexins. By using pull-down assays, here we show binding of a ubiquitin protein ligase, Nedd4, to the C-terminus of connexin43. This observation was confirmed in vivo by coimmunoprecipitation and immunofluorescence, showing colocalization of Nedd4 and connexin43. Binding of Nedd4 to its interaction partners is generally carried out by its WW domains. Our results indicate that the interaction with connexin43 occurs through all three WW domains of Nedd4. Furthermore, whereas WW1 and WW2 domains mainly interact with the unphosphorylated form of connexin43, WW3 binds phosphorylated and unphosphorylated forms equally. In addition, using the surface plasmon resonance approach we show that only the WW2 domain binds to the PY motif located at the Cterminus of connexin43. Suppression of Nedd4 expression with siRNA resulted in an accumulation of gap junction plaques at the plasma membrane, suggesting an involvement of the ubiquitin protein ligase Nedd4 in gap junction internalization.
protein-protein interaction modules of 38-40 residues in length containing two highly conserved tryptophan residues and a conserved proline residue (Chen and Sudol, 1995;Sudol et al., 1995). According to their ligand preference the WW domains have been grouped into four classes (Kay et al., 2000). Nedd4 WW domains bind ligands containing a PY motif and thus belong to Class-I WW domains (Kasanov et al., 2001).
Cx43 contains a PY motif in its C-terminus, which is flanked by phosphorylation sites at Ser279 and Ser282 and is degraded in part by the ubiquitylation pathway (Laing and Beyer, 1995;Rutz and Hulser, 2001). This, and the fact that phosphorylation of Cx43 plays an important role in channel gating, suggests that Cx43 is also targeted by ubiquitin protein ligases in a similar form as the epithelial Na + channel proteins (Asher et al., 2003). However, no ubiquitin protein ligases have hitherto been reported to interact with any of the connexins, and the molecular details of connexin ubiquitylation and degradation are totally unknown.
Here we show that Cx43 binds both in vitro and in vivo the ubiquitin protein ligase Nedd4. We also demonstrate that all three WW domains of the rat Nedd4 interact with rat Cx43 and that this interaction is modulated by the phosphorylation state of Cx43. Further, we prove that the Nedd4 domain WW2 binds to the PY motif at the C-terminus of Cx43, whereas domains WW1 and WW3 bind different motifs in Cx43, so far not identified. Finally, depletion of Nedd4 by siRNA results in the accumulation of gap junction plaques at the plasma membrane, suggesting that Nedd4 plays a role in the internalization/ degradation process of Cx43.

Results
The C-terminal region of Cx43 binds to Nedd4 To identify cellular proteins interacting with Cx43 we used glutathione S-transferase (GST) fusion proteins of the Cx43 intracellular C-terminus (Cx43 CTer) and N-terminus (Cx43 NTer) (Fig. 1A) and performed pull-down assays with WB-F344 cell extracts. As shown in Fig. 1B, we detected several bands bound in the samples containing cell lysates but absent in the control samples containing buffer instead of lysate. To identify the bound proteins, the bands were eluted and analyzed by nanoelectrospray ionization mass spectrometry (nanoESI-MS). Three of the analyzed proteins could unambiguously be identified by computer analysis using the search engine Mascot (Matrix Science, London, UK), all of them bound to Cx43 CTer. One protein was identified as Nedd4, an ubiquitin protein ligase that is known to mediate ubiquitylation-dependent downregulation of integral membrane proteins (Staub et al., 1997;Schwake et al., 2001). The other two proteins, zonula occludens-1 and zonula occludens-2 were already known to interact with the Cterminal tail of Cx43 in GST pull-down assays (Toyofuku et al., 1998;Giepmans et al., 2001a;Singh and Lampe, 2003). These results therefore validate the approach used and identify Nedd4 as a new interaction partner of Cx43.
To substantiate these results, the pull-down experiments were repeated and Nedd4 was identified by immunoblotting with an anti-rat Nedd4 antibody. As shown in Fig. 1C, the Cterminal tail of Cx43 binds a protein with the same molecular weight as Nedd4, and that reacts with the anti-Nedd4 antibody. No reacting band could be detected when the N-terminus of Cx43 or the GST moiety were used for the pull-down assays.
These results demonstrate that the C-terminal region of Cx43 interacts with Nedd4 in vitro.
Full-length Cx43 and Nedd4 interact in vivo and colocalize in WB-F344 cells To confirm our in vitro results, we investigated the interaction of Cx43 with Nedd4 in vivo by using coimmunoprecipitation as well as double immunofluorescence microscopy. To analyze the endogenous Cx43-Nedd4 interaction, anti-Cx43 antibody (71-0700) immunoprecipitates were prepared from WB-F344 cells and the presence of Nedd4 in the immunoprecipitates was investigated by immunoblotting. As shown in Fig. 2A (lower panel), Nedd4 could be clearly identified in Cx43 immunoprecipitates, thus demonstrating that full-length Cx43 is also able to bind Nedd4 under in vivo conditions.
Interestingly, a thorough analysis of the results revealed that the anti-Cx43 antibody 71-0700 exclusively immunoprecipitates the phosphorylated forms P1 and P2 of Cx43, whereas the non-phosphorylated P0 form could not be pulled down with this antibody ( Fig. 2A; upper panel). These results demonstrate that Nedd4 is able to bind P1 and/or P2 phosphorylated isoforms of Cx43 although it cannot be excluded that Nedd4 also interacts with non-phosphorylated Cx43. The interaction between Cx43 and Nedd4 was further substantiated by immunofluorescence analyses. Labelling of WB-F344 cells with anti-Cx43 antibody C13720 (Fig. 2Ba) showed gap junctional staining at cell-cell interfaces, as well as a faint staining of the perinuclear region. Further, Nedd4 was distributed throughout the cell with no apparent nuclear staining being visible (Fig. 2Bb). The merged images showed colocalization of both proteins (Fig. 2Bc, yellow) in structures that in some cases were associated with the plasma membrane and in other cases were intracellular (see inset). Taken together, these results suggest an interaction of Cx43 and Nedd4 under in vivo conditions, and also suggest that this interaction remains, at least temporarily, upon gap junction internalization.

Nedd4 domains WW1, WW2 and WW3 bind to Cx43
The rat Nedd4 protein (rNedd4) contains three WW domains, protein-protein interaction modules of 38-40 amino acids that mediate binding to proline-rich PY motifs (Staub et al., 1996;Fotia et al., 2004;Ingham et al., 2004). To determine whether any of the three WW domains of rNedd4 interacts with Cx43, GST-fusion proteins of individual WW domains from rNedd4 were prepared. Lysates of WB-F344 cells were incubated with GST-fusion proteins containing the different rNedd4 WW domains, or with GST alone as a control. Samples that were incubated with buffer instead of cell lysate were used as further negative controls. Subsequent immunoblotting with anti-Cx43 was used to monitor protein binding. The results of this experiment are shown in Fig. 3A (upper panel). All three WW domains of rNedd4 bound to rat Cx43 (rCx43). Whereas WW2 mainly interacted with non-phosphorylated Cx43 (P0; arrow), WW3 bound nearly equal amounts of both non-phosphorylated (P0; arrow) and phosphorylated Cx43 (P1; arrow). Domain WW1 of rNedd4 seemed to bind preferentially the nonphosphorylated form of Cx43 (P0; arrow). But it cannot be excluded that domain WW1 is also able to interact with phosphorylated Cx43, because the amount of detected Cx43 Journal of Cell Science 119 (17) bound to the WW1 domain is much lower than the amount of Cx43 in the WW2 and WW3 precipitates. Furthermore, the phosphorylated forms of Cx43 seemed to bind at a much lower level to both WW2 and WW3 domains than the nonphosphorylated isoform.
These data support the direct binding of Nedd4 to the Cx43 protein already shown in Fig. 1B,C and Fig. 2A, and further suggests that WW1, WW2 and WW3 are responsible for the interaction of rNedd4 and rCx43.
To test whether these bindings are dependent on the presence of growth factors, we performed pull-downs on serum-starved WB-F344 cells. Under these conditions, domains WW1, WW2 and WW3 bound to the non-phosphorylated P0 isoform of Cx43 (Fig. 3A, arrows in lower panel). After treating serum-starved cells with EGF, a marked increase in the phosphorylated forms of Cx43 was observed, whereas the nonphosphorylated form completely disappeared when the blots were incubated with antibody 71-0700 detecting nonphosphorylated and several phosphorylated forms of Cx43 ( Fig. 3B; EGF in upper panel). In EGF-treated WB-F344 cells, all three WW domains of rNedd4 bound to phosphorylated rCx43 molecules. Consistent with the affinity of the WW3 domain to both non-phosphorylated and phosphorylated Cx43 ( Fig. 3A; upper panel), the increase in Cx43 phosphorylation induced by EGF treatment led to an increase in the amount of phosphorylated Cx43 bound to the WW3 domain ( Fig. 3B; upper panel). Under these conditions, domain WW3 bound ~10 times more Cx43 than domain WW2. Independently of the presence of EGF in the medium, the binding capacity of the WW1 domain to Cx43 was 5-10 times lower than that of domain WW2 (Fig. 3A, upper panel of B).
Previous studies have shown that EGF-induced activation of the Ras-Raf-MEK-MAPK signal-transduction pathway leads to MAPK-mediated phosphorylation on Cx43 at serine residues 255, 279 and 282 (Warn-Cramer et al., 1996;Warn-Cramer et al., 1998). The anti-Cx43 antibody 71-0700 used for the above experiments recognizes in immunoblots the non-phosphorylated The anti-Cx43 antibody 71-0700 was used to immunoprecipitate protein complexes in lysates of WB-F344 cells. Cx43 and the associated Nedd4 were detected in the immunoblot with an anti-Cx43 antibody (71-0700) and anti-Nedd4 antibody, respectively. P1 and P2 label putatively different phosphorylation forms of Cx43; P0 indicates the non-phosphorylated form. (B) Colocalization of endogenous Cx43 and Nedd4 in plasma-membraneassociated and intracellular structures. Cultures of WB-F344 cells were double labelled for Cx43 (green; a,c,d) and Nedd4 (red;b,c,d). Cx43 and Nedd4 are colocalized at appositional membranes and within discrete intracellular structures (c, arrows; yellow). This distribution is evidenced in the enlarged part of c, shown in the inset (d, arrows). Bars, 20 m. and several phosphorylated forms of Cx43 but not the S279/S282 phosphorylated forms (where both serine residues are phosphorylated). To specifically detect the S279/S282 phosphorylated forms of Cx43, we developed the polyclonal antibody SA226P (Leykauf et al., 2003). As expected, EGFinduced S279/S282 phosphorylation of Cx43 was detected with antibody SA226P (Fig. 3B, EGF in lower panel). An interaction between domain WW2 and the S279/S282 phosphorylated forms of Cx43 was detected by a faint band in the immunoblot using antibody SA226P. However, no binding of S279/S282phosphorylated Cx43 to the rNedd4 domains WW1 and WW3 was detectable (Fig. 3B, lower panel). The low affinity of the WW1 domain to non-phosphorylated and phosphorylated Cx43 led to low reactivities in the immunoblots (Fig. 3A, upper panel of B). The amount of pulled down Cx43 might therefore be below the detection level in the lower panel of Fig. 3B. Thus, it cannot be excluded that domain WW1 was also able to bind S279/S282 phosphorylated Cx43.
These results indicate that rNedd4 binds Cx43 via its domains WW1, WW2 and WW3 and suggest that the binding of both proteins is modulated by the phosphorylation state of Cx43. This is supported by the finding that S279/S282 phosphorylated Cx43 showed no binding to rNedd4 domains WW1 and WW3, whereas EGF-mediated phosphorylation of Cx43 at residues other than S279 and S282 increased the binding capacity of domain WW3 to Cx43.
Only the WW2 domain of Nedd4 binds to the PY motif of Cx43 WW domains have been grouped into four classes according to their ligand preferences (Kay et al., 2000). The WW domains of Nedd4 belong to Class-I WW domains, which are characterised by binding to ligands containing an xPPxY motif (PY motif) (Kasanov et al., 2001). As shown in Fig. 4A amino acids 282-286 at the C-terminal region of Cx43 contain a putative PY motif (marked in red). Two serine residues located N-terminally and within the PY motif at positions 279 and 282 are phosphorylated by MAPK resulting in a rapid and transient disruption of gap-junction communication (Warn-Cramer et al., 1996;Warn-Cramer et al., 1998). Interestingly, phosphorylated residues near or within the PY motif in other proteins seem to play an additional role in binding the WW domains of Nedd4 (Shi et al., 2002;Ingham et al., 2004). In order to determine the importance of the PY motif for the interaction between Cx43 and Nedd4 and to definitely ascertain whether phosphorylation of S279 and S282 has a significant effect on binding, we used surface plasmon resonance (SPR). Two N-terminal biotinylated peptides encompassing the PY motif of Cx43 (Fig. 4A) were immobilized on the streptavidin surface of a sensor chip. One of the peptides was synthesized with the S279 and S282 in the non-phosphorylated state (Cx43 CT) whereas the other peptide was phosphorylated at both serine residues (Cx43 CT phosph ). As a control for non-specific binding, a biotinylated peptide corresponding to the Nterminus of Cx43 was used (Cx43 NT; Fig. 4A). WW domain binding affinities were measured using GST-fusion proteins of the three individual WW domains from rNedd4. Neither GST-WW1 nor GST-WW3 bound to any of the immobilized peptides (data not shown). By contrast, GST-WW2 bound to the phosphorylated (Fig. 4C) as well as to the nonphosphorylated peptide of the Cx43 C-terminus (Fig. 4D), in a concentration-dependent manner. The affinity of GST-WW2 was slightly higher for the phosphorylated peptide than the non-phosphorylated peptide (Fig. 4C,D). The low off-rates indicate a stable binding. As expected, no binding of GST- Fig. 3. Rat Nedd4 domains WW1, WW2 and WW3 bind to Cx43. The first lanes were loaded with cell lysates of WB-F344 cells as a Cx43 size control. As negative controls for the pull-down analyses GST without any fusion protein was used. Further negative controls were samples of the different GST-WW fusion proteins that were incubated with buffer (-) instead of WB-F344 cell lysates (+). (A) SDS-PAGE and immunoblot of Cx43 obtained in pull-down assays with GST-fusion proteins of rat Nedd4 domains WW1, WW2 and WW3. Cx43 was recognized by anti-Cx43 antibody 71-0700, which detects non-phosphorylated and several phosphorylated forms of Cx43. Upper panel: in pull-down experiments with cell lysates from untreated WB-F344 cells WW1, WW2 and WW3 of rNedd4 bound to rCx43. As indicated by arrows, WW1 and WW2 mainly bound the non-phosphorylated form of Cx43 (P0), whereas domain WW3 interacted with the non-phosphorylated form of Cx43 (P0) and the P1-isoform of Cx43. Lower panel: in the absence of serum (-FCS) all three WW domains of rNedd4 exclusively bound to the non-phosphorylated form of Cx43 (P0; arrows). (B) To strongly increase the phosphorylated forms of Cx43 before performing pull downs with GST-WW1, GST-WW2 and GST-WW3, confluent serum-starved (-FCS) WB-F344 cells were treated with EGF (+EGF). Bound Cx43 were then analyzed by SDS-PAGE and immunoblotting. The second lanes were loaded with cell lysates of EGF-treated WB-F344 cells as a size control for phosphorylated Cx43 molecules. Upper panel: incubation with anti-Cx43 antibody 71-0700 resulted in a strong signal for Cx43 in the WW2 and WW3 precipitates. Under these conditions, domain WW3 bound ~10 times more Cx43 than domain WW2. Although binding of the WW1 domain to Cx43 was detectable in the immunoblot, the binding capacity of the WW1 domain was 5-10 times lower than that of domain WW2. Lower panel: with antibody SA226P, that specifically recognises the S279/S282 phosphorylated forms of Cx43, only a faint band in the WW2 precipitate was detectable, whereas no binding of S279/S282 phosphorylated Cx43 to the rNedd4 domains WW1 and WW3 was visible. P0 indicates the non-phosphorylated Cx43 molecules. P1 and P2 label putatively different phosphorylation forms of Cx43.
WW2 was observed to the peptide of the Cx43 N-terminus (Cx43 NT; Fig. 4B): all signals were indistinguishable from that of the baseline. Determination of K Dapp by steady-state kinetics (Fig. 5A) revealed K Dapp values of around 0.6 M for the phosphorylated peptide and around 1.1 M for the nonphosphorylated peptide of the Cx43 C-terminus. To confirm the specificity of the observed binding, competition experiments were performed, where the GST-WW2 fusion protein was incubated with increasing concentrations of nonbiotinylated phosphopeptides of the Cx43 C-terminus before its injection. Binding of GST-WW2 to the immobilized Cx43 CT phosph peptides decreased in a biphasic fashion with increasing amounts of competing peptides (Fig. 5B), thus Journal of Cell Science 119 (17) confirming the specific binding of the GST-WW2 to the PY motif containing phosphorylated peptide of Cx43 (Cx43 CT phosph ). Together, these experiments demonstrate that the WW2 domain of rNedd4 binds to the peptides containing the PY motif of Cx43 and that binding occurs both with and without phosphorylation of serine residues 279 and 282.

Nedd4 siRNA increases membrane localization of Cx43
To investigate the functional significance of the interaction between Cx43 and Nedd4, we suppressed the expression of Nedd4 using siRNA for Nedd4 and examined its effect on Cx43. WB-F344 cells were transiently transfected with Nedd4 siRNA and the expression level of endogenous Nedd4 was

GST-WW2 [µM]
[RU] Resp. Diff. examined 144 hours after the transfection (Fig. 6). Immunoblotting showed ~75% decrease in the amount of Nedd4 whereas mock transfection and transfection of nontargeting siRNAs had no effect when compared with control cells (Fig. 6A, upper panel). The signals for total Cx43 and for actin differed in a range of ~10%. Actin was used as a loading control (Fig. 6A, lower panel). The comparable expression levels of total Cx43 in cells in the presence and the absence of Nedd4 siRNA (Fig. 6A, middle panel) suggest that Nedd4 is rather involved in the internalization of gap junctions than in the direct degradation of Cx43 from the plasma membrane. Therefore, to analyze whether the cellular distribution of Cx43 was modified in Nedd4 siRNA-treated WB-F344 cells, we performed double immunofluorescence using antibodies against Cx43 and Nedd4. WB-F344 cells transfected with nontargeting siRNA showed structures of gap junction plaques at the plasma membranes between adjacent cells (Fig. 6Ba). Nedd4 signals were found throughout the cells with the exception of the nuclei (Fig. 6Bb). The depletion of Nedd4 by siRNA seemed to increase the amount of Cx43 associated with the plasma membrane, showing a line of typical gap junction plaque structures (Fig. 6Bd). That this effect could be due to the absence of the ubiquitin protein ligase Nedd4 was further supported by the finding that in single cells still expressing some Nedd4 (red cells in Fig. 6Be,Bf), fewer gap junction plaques were found at sites of appositional membranes (arrows in Fig. 6Bd,Bf). The observation that the amount of gap junction plaques at cell-cell interfaces between adjacent Nedd4-expressing cells was lower compared with cells lacking Nedd4 is in agreement with the staining pattern of Cx43 in cells transfected with non-targeting siRNAs (Fig. 6Ba-c).

Nedd4 binds all detectable isoforms of Cx43
In this study we have identified the ubiquitin protein ligase Nedd4 as an interaction partner of Cx43. Cx43 differs from many other transmembrane proteins by its high turnover rate. Both the lysosome and the ubiquitin proteasome machinery have been shown to be involved in the degradation process of Cx43 gap junctions (Larsen and Hai-Nan, 1978;Naus et al., 1993;Laing and Beyer, 1995;Laing et al., 1997;Musil et al., 2000;Rutz and Hulser, 2001). In most cases ubiquitylation is essential for proteasomal degradation. Additionally, it has become apparent that ubiquitylation also plays an important role in the endocytosis of many transmembrane proteins leading to lysosomal degradation (Bonifacino and Weissman, 1998;Rotin et al., 2000;Ingham et al., 2004). This led to speculation about the identity of the protein that transfers ubiquitin onto Cx43 to mark Cx43 either for the lysosome or the proteasome.
Our results demonstrate binding of Nedd4 to the C-terminus of Cx43 in a GST pull-down approach using the rat cell line WB-F344 and subsequent nanoESI-MS analysis (Fig. 1B,C). In addition, our coimmunoprecipitation experiments proved that the interaction between endogenous Cx43 and Nedd4 also occurs in vivo ( Fig. 2A), thus indicating that this binding is a physiological event in the internalization and/or degradation process of Cx43. Furthermore, the results demonstrate that Nedd4 binds to the P1 and/or P2 phosphorylated isoforms of Cx43. Nevertheless, no unambiguous conclusion can be drawn regarding the interaction between Nedd4 and nonphosphorylated Cx43 P0, because this form of Cx43 could not be immunoprecipitated with the anti-Cx43 antibody 71-0700. However, in the pull-down experiments, all three WW domains showed binding to non-phosphorylated Cx43 (Fig. 3A). Taken together, these results prove that the ubiquitin protein ligase Nedd4 binds Cx43 molecules with different phosphorylation states.
rNedd4 domains WW1, WW2 and WW3 prefer different motifs of Cx43 PY motifs of several different ion channels have been shown to act as ligands for WW-domain-containing ubiquitin protein ligases, leading to their ubiquitylation-dependent removal from the plasma membrane (Staub et al., 1996;Staub et al., 1997;Abriel et al., 2000;Schwake et al., 2001). The C-terminal tail of Cx43 contains a putative PY motif (amino acids 282-286) known to be important for degradative processes (Chen and Sudol, 1995). Our results with pull-down experiments demonstrated that domains WW1, WW2 and WW3 of rNedd4 bind to Cx43. In all pull-down experiments the WW1 domain showed a five to ten times reduced binding capacity compared with the binding of domain WW2 to Cx43, suggesting that the WW1 domain of rNedd4 plays a minor role in the interaction of both proteins.
The phosphorylation of Cx43 is not indispensable for the interaction between Nedd4 and Cx43. The Cx43 phosphorylation state rather may modulate the binding of both proteins, because differential binding patterns of Nedd4 to several Cx43 isoforms could be observed in untreated, serumstarved and EGF-treated cells. The specific phosphorylation of Cx43 at S279 and S282 seemed to inhibit the interaction between Cx43 and rNedd4 domain WW3 and possibly that of domain WW1 (Fig. 3A,B).
In untreated cells the WW3 domain bound to equal amounts of non-phosphorylated and phosphorylated Cx43, whereas domain WW2 mainly interacted with the non-phosphorylated form of Cx43. Furthermore, the EGF-induced increase in the amount of phosphorylated Cx43 resulted in a significantly stronger binding signal of phosphorylated Cx43 to domain WW3 compared with that of domains WW1 and WW2. Together, these results indicate that rNedd4 domain WW3 has a higher affinity to phosphorylated Cx43 than domains WW1 and WW2. This raises the question whether the WW3 domain is the predominant mediator of the interaction of rNedd4 to Cx43 at the plasma membrane, where Cx43 is generally phosphorylated (Musil et al., 1990b). This hypothesis is in accordance with the findings that an EGF-mediated shift in the phosphorylation state of Cx43 correlates with an increase in the ubiquitylation state of Cx43 at the plasma membrane, further associated with accelerated Cx43 internalization and degradation (Leithe and Rivedal, 2004a;Leithe and Rivedal, 2004b). Two putative phosphorylation sites of the EGF-induced and activated MAPK (S279 and S282) flank the PY motif of Cx43 on the N-terminal side. Using antibody SA226P, which specifically recognises the S279/S282 phosphorylated forms of Cx43 (Leykauf et al., Journal of Cell Science 119 (17) 2003), we have shown that S279/S282 phosphorylated Cx43 does not bind to domain WW3 of rNedd4 (Fig. 3B). Therefore, EGF-dependent phosphorylation of residues other than S279 and S282 appears to be significant for an interaction of Cx43 with rNedd4 domain WW3.
The results of the SPR approach are concordant with the results of the pull-down assays, because domains WW1 and WW3 failed to bind the PY motif at the C-terminus of Cx43. This suggests that both domains prefer different proline-rich motifs of Cx43 for binding. In this sense, two other types of WW-binding motifs, different from the PY motif, have been described: the PGM motif (Lehman et al., 1998) and the PPLP motif (Chan et al., 1996;Bedford et al., 1997). Whether one of these binding motifs or another motif at the C-terminal tail of Cx43 is responsible for the interaction of Cx43 with domains WW1 and WW3 of rNedd4 remains to be elucidated.

WW2 domain of rNedd4 exclusively binds to the PY motif of Cx43
Our results indicate that WW2 is the only rNedd4 WW domain that binds robustly to the extended PY motif of Cx43. Binding affinity was calculated by fitting steady-state values to simple binding kinetics. The K Dapp value obtained was in the range of 1 M (Fig. 5A). This is in agreement with the values measured for the interaction of mNedd4-1 and mNedd4-2 with other membrane proteins, like the epithelial Na + channel (Asher et al., 2001;Shi et al., 2002;Asher et al., 2003). Phosphorylation of S279 and S282 at the N-terminally extended PY motif almost doubled the binding affinity, an effect that was only weakly statistically significant (P<0.1). Nevertheless, this value is comparable to values obtained from binding of the serinephosphorylated PY motif in the ␤ subunit of the epithelial Na + channel to rNedd4 WW4 domain (Henry et al., 2003). Taken together, the results of the pull-down experiments and the SPR approach showed that the interaction of Cx43 and Nedd4 is at least ensured by the binding of the extended PY motif of Cx43 to domain WW2 of Nedd4, and that phosphorylation of amino acids S279 and S282 is not required for this binding.
Thus, our results indicate that the interaction between Nedd4 and Cx43 occurs involving multiple sites of both proteins, from which only the WW2 domain binds the canonical PY motif of Cx43, whereas domains WW1 and WW3 bind different, as yet unknown motifs.
Nedd4 interacts with Cx43 and seems to trigger its internalization By means of a siRNA approach, we have provided direct experimental evidence for a role of Nedd4 in the organization of gap-junction plaques in WB-F344 cells. Whereas suppressing the expression of Nedd4 had no influence on the level of total Cx43 in the cell, the absence of Nedd4 seemed to lead to an accumulation of gap junctions in the plasma membrane when compared with Nedd4-expressing cells (Fig.  6B). This is further confirmed by the results of the double immunofluorescence under normal conditions, where colocalization of Nedd4 and Cx43 was detected at the plasma membrane as well as in intracellular structures (Fig. 2B). Our results are in accordance with those published by Rutz and Hulser, who show ubiquitin-conjugated Cx43 at gap junction plaques (Rutz and Hulser, 2001). These observations indicate, as it is the case with ion channels, that Nedd4 binds to and ubiquitylates Cx43 molecules at the cell surface, leading to their internalization probably via the interaction with accessory proteins (Abriel et al., 2000;Debonneville et al., 2001;Snyder et al., 2002;Henke et al., 2004). Furthermore, the colocalization of Nedd4 and Cx43 in intracellular structures suggests that the protein complex is stable, at least temporarily, after internalization of the gap junctions.
The general picture emerging from our experiments is that Nedd4 binds in vitro and in vivo to Cx43 through multiple interactions and that these interactions are partially modulated by phosphorylation. It has been shown that Cx43 possesses at least 14 amino acids susceptible to phosphorylation (Lampe and Lau, 2004). Since no antibodies are presently available recognizing the protein specifically phosphorylated at different amino acids, a comprehensive description of the events correlating specific Cx43 phosphorylation and Nedd4 binding is currently not possible.
In summary, the ubiquitin protein ligase Nedd4 binds in vivo to Cx43 localized in gap junctions, and this interaction is mediated by domains WW1, WW2 and WW3 of rNedd4, and by the PY motif at the C-terminus of rat Cx43. The interaction of both proteins plays an explicit role in the internalization of gap junctions and posterior degradation of the gap-junction proteins.

Materials and Methods
Cell culture and antibodies WB-F344 rat liver epithelial cells were cultured in DMEM (GIBCO) supplemented with 10% fetal calf serum and penicillin-streptomycin. Before treatment, WB-F344 cells were grown to confluency and serum-starved for 24 hours. Cells were treated with 100 ng/ml EGF for 15 minutes. Three different anti-Cx43 antibodies were used. In immunoblots, Cx43 was detected with a rabbit polyclonal antibody generated against a peptide corresponding to a segment of the third cytoplasmic domain of rat Cx43 (71-0700, Zymed Laboratories). A mouse monoclonal antibody that was raised against amino acids 252-270 of the Cx43 sequence was used for immunoflourescence (C13720, Transduction Laboratories). Both antibodies detect the non-phosphorylated and several phosphorylated but not the S279/S282 phosphorylated forms of Cx43 (Nagy et al., 1997;Giepmans et al., 2001b). SA226P is a polyclonal rabbit antibody that specifically detects the S279/S282 phosphorylated forms of Cx43, as described previously (Leykauf et al., 2003). In immunoflourescence and in immunoblots Nedd4 was detected by a rabbit polyclonal antibody generated against the WW2 domain of rat Nedd4 (Upstate Biotechnology).

Plasmid construction
Forward primers with BamHI cleavage sites at their 5Ј ends and reverse primers with EcoRI cleavage sites at their 5Ј ends were used to amplify the N-terminus (Cx43 NTer) and the C-terminal tail of Cx43 (Cx43 CTer). The polymerase chain reaction products were digested with BamHI and EcoRI and ligated into the bacterial expression vector pGEX-4T-3 (Pharmacia Biotech), previously digested with BamHI and EcoRI. These constructs produce fusion proteins of Cx43 NTer and Cx43 CTer, respectively, C-terminally bound to glutathione S-transferase (GST). Plasmid constructs for the expression of GST-WW1, GST-WW2 and GST-WW3 fusion proteins (pGEX-2T-K-WW1, pGEX-2T-K-WW2, pGEX-2T-K-WW3) were kindly provided by Dr D. Rotin (Hospital for Sick Children, Toronto, Canada).

GST pull downs
The GST-fusion plasmids and the empty pGEX-4T-3 vector were transformed into Escherichia coli strain BL21. Bacteria were grown overnight, 5 ml were transferred to 500 ml of fresh LB, and incubated at 37°C until an OD 600 0.5-0.6 was reached. Expression was induced by addition of 0.1 mM Isopropyl-beta-Dthiogalactopyranoside (Sigma) and incubation at 37°C for 1 hour. The bacterial pellet was collected by centrifugation and resuspended with PBS. Cells were lysed by sonication and the extract was cleared by centrifugation. To prepare fusionprotein-coated beads, the supernatant was bound to glutathione-Sepharose 4B beads (Amersham Pharmacia). The desired proteins coupled to beads (GST, GST-Cx43 NTer, GST-Cx43 CTer, GST-WW1, GST-WW2 and GST-WW3) were resuspended in E1A buffer (250 mM NaCl, 0.1% Triton X-100, 10 mM MgCl 2 , 50 mM HEPES pH 7.4). 6.7 mg E1A lysates of WB-F344 cells were incubated with GST-fusion proteins coupled to glutathione-Sepharose 4B overnight at 4°C on a rotating wheel. Bound proteins were eluted with SDS-PAGE sample buffer and subjected to SDS-PAGE. Gels were stained with Coomassie SimplyBlue TM SafeStain (Invitrogen) and the bands were excised and analyzed using nanoelectrospray ionization mass spectrometry.

Surface plasmon resonance
Real-time biomolecular interaction analyses were performed with a Biacore 3000 surface plasmon resonance biosensor. Two assays were used to investigate the properties of an 18 aa Cx43 peptide spanning the C-terminal PY motif of Cx43. First, the direct binding of the three WW domains of Nedd4 to the phosphorylated and non-phosphorylated peptide was measured. Second, competition assays using free, unbiotinylated peptides were performed on the same system.
For kinetics experiments N-terminal biotinylated peptides (biotin-CSSPTAPLpSPMpSPPGYKL; biotin-CSSPTAPLSPMSPPGYKL) containing the phosphorylated and non-phosphorylated PY motif of Cx43, respectively, were immobilized on a sensor chip with a streptavidin surface, that is able to capture biotinylated molecules. As negative control, a biotinylated 18 aa Cx43 peptide of the cytoplasmic N-terminus of Cx43 (biotin-SALGKLLDKVQAYSTAGG) was used. 40 l of solutions containing between 25 and 50 ng/ml of the different biotinylated peptides were injected resulting in binding level responses from 150 to 200 resonance units (RU). To test their binding to the prepared peptide surface the GST-WW fusion proteins and recombinant GST (45 l) were injected. The tested protein concentrations ranged between 48.0 nM and 3.0 M. All experiments were carried out at room temperature with Biacore HBS-EP buffer (150 mM NaCl, 3 mM EDTA, 0.005% surfactant P20, 10 mM HEPES, pH 7.4) as running buffer. The flow rate was 15 l/minute. In control experiments, a biotinylated 18 aa peptide derived from human albumin (biotin-AMLSLGTKADTHDEILEG) was used to evaluate non-specific binding. For competition assays, the unbiotinylated phosphopeptide (CSSPTAPLpSPMpSPPGYKL) was diluted into a solution containing GST-WW2 fusion proteins at a final concentration of 380 nM. After incubating at room temperature for at least 15 minutes the samples covering the concentration range of phosphopeptide between 98 nM and 50 M were injected.

Coimmunoprecipitation analysis
Cell lysates of confluent WB-F344 cells were prepared in modified RIPA buffer. Protein concentration was determined by the Bio-Rad DC protein assay. 600 g of the lysate were incubated with 2 g of polyclonal anti-Cx43 antibody (71-0700) bound to protein G-Sepharose for 4 hours at 4°C in Cx43 immunoprecipitation buffer (150 mM NaCl, 1.5 mM MgCl 2 , 1 mM EGTA, 10% glycerol, 1% Triton X-100, 50 mM HEPES, pH 7.5). Samples were subjected to western blot analysis with anti-Nedd4 WW2 antibodies. Peroxidase-conjugated secondary antibodies against rabbit IgG antibodies and enhanced chemiluminescence (Amersham Pharmacia) were used to visualize primary antibody-antigen complexes. After stripping with Re-Blot Plus Strong Solution (Chemicon) the membrane was incubated with anti-Cx43 antibodies (71-0700).

Immunofluorescence
Treated or untreated WB-F344 cells grown on glass coverslips were fixed with icecold methanol and acetone and incubated with the monoclonal antibody anti-Cx43 (C13720) and the polyclonal antibody anti-Nedd4 WW2. After washing in PBS, cells were incubated with a goat anti-mouse antibody coupled to Alexa Fluor 488 and a goat anti-rabbit antibody coupled to Alexa Fluor 594. Coverslips were again washed and mounted in antifading medium. Pictures were taken with a confocal microscope (Leica DM IRBE).

siRNA-mediated knockdown
siRNA targeting murine Nedd4 (M-101184-00) and a non-targeting control sequence (D-001210-01-05) were obtained from Dharmacon, WB-F344 cells were grown on 12-well plates until they reached 80% confluency. siRNA was complexed with Lipofectamine 2000 (Invitrogen) according to the manufacturer's recommendations. The resulting formulations were applied to each well at a final concentration of 200 pmol/well. After 18 hours incubation at 37°C, transfected cells were trypsinized and 8ϫ10 5 cells were transferred onto 12-well plates with or without glass coverslips. 144 hours after transfection, cells were used for western blot analysis and immunofluorescence, respectively.