Ubiquitylation of the gap junction protein connexin-43 signals its trafficking from early endosomes to lysosomes in a process mediated by Hrs and Tsg101

Gap junctions are arrays of intercellular channels that enable neighbouring cells to communicate directly by exchanging ions, signalling molecules and small metabolites (Saez et al., 2003). Gap junction channels are composed of two channel structures called connexons, each of which are contributed by the two adjacent cells. Connexons consist of six connexin proteins. There are 21 known members of the connexin protein family in the human genome, of which the best-studied isoform is connexin-43 (Cx43; also known as Gja1) (Sohl and Willecke, 2003). Connexins play important roles in the regulation of cell growth and tissue homeostasis, and dysfunctional intercellular communication via gap junctions has been implicated as a causative factor in heart failure, neuropathology, deafness, skin disorders and cataracts (Wei et al., 2004). There is also significant evidence that loss of gap junctional communication is an important step in carcinogenesis (Leithe et al., 2006b; Mesnil et al., 2005).

Gap junctions are dynamic plasma membrane domains. Newly synthesized connexons are continually added to the edges of existing gap junctions and dock with connexons in the adjacent cells to form functional intercellular channels (Gaietta et al., 2002; Lauf et al., 2002). Following their assembly into gap junctions, Cx43 moves toward the centre of the gap junction, where it is removed by endocytosis (Gaietta et al., 2002; Lauf et al., 2002). Cx43 has a high turnover rate in most tissues, with a half-life of 1.5-5 hours (Fallon and Goodenough, 1981; Laird et al., 1991). Cx43 is known to be tightly regulated by phosphorylation (Solan and Lampe, 2009). Among the kinases involved in phosphorylation of Cx43 are protein kinase C (PKC) and mitogen-activated protein (MAP) kinase (Lampe et al., 2000; Warn-Cramer et al., 1998). Modulation of connexin degradation has been suggested to be an important mechanism by which cells regulate the level of gap junctional intercellular communication (Berthoud et al., 2004; Laird, 2005; Musil et al., 2000; Thomas et al., 2003). However, the molecular machinery involved in mediating connexin degradation has remained poorly understood.

During endocytosis of gap junctions, both membranes of the junction are internalized into one of the adjacent cells and thereby form a double-membrane vacuole called an annular gap junction or connexosome (Jordan et al., 2001; Larsen and Hai, 1978; Leithe et al., 2006a; Nickel et al., 2008; Piehl et al., 2007). Following internalization of gap junctions, connexins are degraded in lysosomes (Laing et al., 1997; Leithe and Rivedal, 2004b; Leithe et al., 2006a; Naus et al., 1993; Qin et al., 2003; VanSlyke et al., 2000; Vaughan and Lasater, 1990). Based on immunoelectron microscopy studies, we have previously proposed that the internalized, annular gap junction undergoes a maturation process from a double membrane vacuole to a multivesicular endosome with a single limiting membrane (Leithe et al., 2006a). This processing of the annular gap junction was found to be associated with trafficking of Cx43 to early endosomes, prior to its degradation in lysosomes (Leithe et al., 2006a). The observation that the gap junction double membrane is processed into two single membranes after its internalization implies that the gap junction channels undock during or shortly after internalization. Thus, in this scenario for Cx43 degradation, most Cx43 localized in early endosomes is expected to exist as undocked connexons. The early endosome is a major sorting station for proteins internalized from the plasma membrane (Raiborg and Stenmark, 2009). After entering the early endosome, endocytosed proteins can be trafficked further along the degradation pathway to the lysosome, be recycled to the plasma membrane, or be transported to the trans-Golgi network. By orchestrating the trafficking of endocytosed proteins, early endosomes play important roles in regulating the levels of proteins at the plasma membrane (Gould and Lippincott-Schwartz, 2009). However, the functional importance of the early endosome in the formation and degradation of gap junctions has not been investigated.

As early endosomes mature into late endosomes, proteins that are destined for lysosomal degradation are incorporated into intraluminal vesicles that bud from the limiting membrane of the endosome (Raiborg and Stenmark, 2009). Subsequently, fusion between late endosomes and lysosomes causes proteolytic degradation of the intravacuolar vesicles and their proteins. Transport of proteins into multivesicular endosomes requires positive sorting signals present at the cytosolic domains of the proteins. Conjugation of ubiquitin is the best-known sorting signal for lysosomal trafficking of endocytosed growth factor receptors (Raiborg and Stenmark, 2009). Ubiquitylated growth factor receptors are recognized by the ubiquitin-binding protein Hrs (hepatocyte growth factor-regulated tyrosine kinase substrate; also known as Hgs), which together with STAM (signal-transducing adaptor molecule) and Eps15b forms a complex at endosomes referred to as ESCRT (endosomal sorting complex required for transport)-0 (Bache et al., 2003; Bilodeau et al., 2002; Hirano et al., 2006; Raiborg et al., 2002; Roxrud et al., 2008). Subsequently, the ubiquitylated receptor is thought to be transferred to the ubiquitin-binding protein Tsg101 (tumor susceptibility gene 101), which is part of a heterotrimeric complex called ESCRT-I (Babst et al., 2000; Bishop and Woodman, 2001; Bishop et al., 2002; Lu et al., 2003). The cargo protein is then transported to the protein complexes ESCRT-II and -III, before its incorporation into intraluminal vesicles (Piper and Katzmann, 2007). Prior to the transport of the receptor into the lumen of the endosome, it is deubiquitylated by isopeptidases recruited by components in the ESCRT-III complex, enabling reuse of ubiquitin (Kato et al., 2000; McCullough et al., 2006; Tanaka et al., 1999). Endocytosed receptors that do not undergo ubiquitylation are not sorted to the lysosome but are instead recycled from the early endosome to the plasma membrane (Raiborg and Stenmark, 2009).

Several lines of evidence indicate that ubiquitin plays a central role in the regulation of Cx43 degradation. Firstly, Cx43 is able to undergo ubiquitylation, and inactivation of the ubiquitin-activating enzyme E1 is associated with increased levels of Cx43 protein (Laing and Beyer, 1995). Secondly, the E3 ubiquitin ligase Nedd4 binds to Cx43 and regulates the level of gap junctions at the plasma membrane (Leykauf et al., 2006). Thirdly, ubiquitylation of Cx43 is strongly induced in response to activation of MAP kinase or PKC (Leithe and Rivedal, 2004a; Leithe and Rivedal, 2004b). However, the functional role of ubiquitin in the degradation of gap junctions remains unknown. In the present study, we show that endocytosis of Cx43 is associated with a strong relocalization of ubiquitin to gap junction plaques. We provide evidence that Cx43 remains ubiquitylated along its trafficking to early endosomes. The data suggest that early endosomes have an essential role in mediating trafficking of Cx43 to lysosomes, in a process mediated by Hrs and Tsg101.

Ubiquitylation of Cx43 in IAR20 cells is efficiently induced by exposure to epidermal growth factor (EGF) or the PKC activator 12-O-tetradecanoylphorbol 13-acetate (TPA) (Leithe and Rivedal, 2004a; Leithe and Rivedal, 2004b). We have previously shown that EGF- and TPA-induced endocytosis of Cx43 gap junctions is associated with loss of the detergent resistance of Cx43 (Sirnes et al., 2008). This loss of the Cx43 detergent resistance was suggested to reflect the separation of the gap junction double membrane during endocytosis, as observed by immunoelectron microscopy (Leithe et al., 2006a). As a first approach to elucidate the role of ubiquitin in the endocytosis of gap junctions, we considered it important to determine the detergent resistance of the ubiquitylated Cx43 pool. Cx43 from IAR20 cells, forms three distinct bands on SDS-PAGE (Fig. 1). The two upper bands (Cx43-P1 and Cx43-P2) represent Cx43 organized in gap junctions and they are Triton X-100 resistant in unstimulated cells (Musil and Goodenough, 1991; Sirnes et al., 2008). As expected, TPA strongly induced transformation of Cx43 into a Triton X-100 soluble fraction (Fig. 1). Interestingly, the ubiquitylated Cx43 pool was found both in the Triton X-100-insoluble and -soluble fractions. These data suggest that ubiquitylated Cx43 exists both in intact gap junctions (i.e. in a Triton X-100-insoluble state) as well as in gap junctions that are in the process of loosing their double membrane structure (i.e. in a Triton X-100-soluble state).

We next aimed to elucidate the association between ubiquitin and Cx43 by confocal immunofluorescence microscopy. For this purpose, we used the FK2 antibody, which specifically detects ubiquitin conjugated to proteins (Fujimuro et al., 1994). In agreement with previous findings, most Cx43 in IAR20 cells was organized as gap junctions between neighbouring cells (Fig. 2A) (Leithe and Rivedal, 2004a; Leithe and Rivedal, 2004b). As expected, ubiquitin was found to localize both in the nucleus and in the cytoplasm (Fig. 2A). Under normal cell growth conditions, ubiquitin was rarely found to colocalize with gap junctions. Importantly, treating the cells with TPA for 15 minutes caused a dramatic relocalization of ubiquitin to Cx43 gap junction plaques (Fig. 2B; supplementary material Fig. S1A). Interestingly, the Cx43 gap junctions that colocalized with ubiquitin were often observed to be in the process of internalizing from the plasma membrane into one of the adjacent cells. Internalization of gap junctions occurred both in the centre and in the periphery of the gap junction plaque. Often, only one or two gap junctions in a cell were found to be in the process of internalizing into the cytosol (Fig. 2C). In these cases, the gap junctions that were in the process of internalizing colocalized with ubiquitin, whereas the other gap junctions did not colocalize with ubiquitin.

Sometimes, annular gap junctions that had completely detached from the plasma membrane (Fig. 2D) or were in the process of being internalized (supplementary material Fig. S1B) were observed. The annular gap junctions were found to colocalize with ubiquitin. Notably, however, most gap junctions that were in the process of internalizing from the plasma membrane did not appear to form annular gap junctions. Instead, the gap junctions that internalized showed a more diffuse Cx43 staining compared to gap junction areas that were not in the process of internalizing (Fig. 2B,C). This diffuse Cx43 staining possibly reflects the separation of the gap junction double membrane and loss of the detergent resistance of Cx43, as described above and in previous studies (Leithe et al., 2006a; Sirnes et al., 2008).

The TPA-induced colocalization between ubiquitin and Cx43 gap junctions was counteracted by the PKC inhibitor GF109203 (Fig. 2E), in accordance with our previous finding that TPA-induced ubiquitylation of Cx43 is mediated by PKC (Leithe and Rivedal, 2004b). Also the MEK1 inhibitor PD98059 strongly counteracted the TPA-induced recruitment of ubiquitin to Cx43 gap junctions (supplementary material Fig. S1C), in agreement with our previous observation that PKC-induced ubiquitylation of Cx43 is partly mediated through the MAP kinase pathway (Leithe and Rivedal, 2004b).

Collectively, these results suggest that ubiquitin colocalizes with Cx43 gap junction plaques in response to PKC activation, and that Cx43 remains ubiquitylated during and after internalization.

In agreement with previous studies, most Cx43 was found to localize in intracellular vesicles following 30 minutes of TPA exposure (Fig. 3A) (Leithe and Rivedal, 2004b). Interestingly, these vesicles were often found to be ubiquitin positive, supporting the notion that Cx43 remains ubiquitylated after internalization. We have previously reported that following internalization of Cx43 gap junctions, Cx43 is trafficked via the early endosome before its degradation in lysosomes (Leithe et al., 2006a). Based on the above data, showing that Cx43 remained ubiquitylated after its internalization from the plasma membrane, we asked whether ubiquitylation of Cx43 could play a role in mediating trafficking of Cx43 from the early endosome to the lysosome. To test this hypothesis, we focused on the ubiquitin-binding proteins Hrs and Tsg101. Both proteins have been reported to interact with ubiquitylated growth factor receptors at the early endosome and are essential for mediating their trafficking to the lysosome (Raiborg and Stenmark, 2009). Interestingly, following 30 minutes of TPA treatment, Cx43 was found to partly colocalize with Hrs and Tsg101 in intracellular vesicles, as determined by confocal microscopy (Fig. 3B,C).

To investigate the functional importance of Hrs and Tsg101 in Cx43 degradation, endogenous Hrs and Tsg101 were depleted by small interfering RNA (siRNA). Following transfection of IAR20 cells with siRNA sequences to Hrs and Tsg101 alone or in combination, the Hrs protein level was reduced by approximately 70% compared with the control, whereas Tsg101 expression was reduced approximately 95%, compared to cells transfected with a control siRNA sequence (Fig. 4A). Depletion of Hrs did not cause major changes in the localization of Cx43 in untreated cells (Fig. 4B). Cells in which Tsg101 was depleted appeared to have increased Cx43 staining and enlarged gap junctions compared with control cells. However, cells depleted of Tsg101 did not show more Cx43 staining in intracellular vesicles than control cells. By contrast, when Hrs and Tsg101 were simultaneously depleted, Cx43 was found to localize both in intracellular vesicles and in gap junctions (Fig. 4B).

We next aimed to elucidate whether Hrs and Tsg101 are involved in PKC-induced degradation of Cx43. In these experiments, TPA was co-incubated with cycloheximide, to block new protein synthesis. As expected, TPA treatment for 1.5 hours caused a strong loss of Cx43 staining in cells transfected with a control siRNA sequence, as determined by immunofluorescence microscopy (Fig. 4B). By contrast, cells in which Hrs or Tsg101 was depleted showed significant staining of Cx43 in intracellular vesicles at this time point. As determined by confocal microscopy, these vesicles were often found to be positive for the early endosome marker EEA1 (Fig. 5). The TPA-induced loss of Cx43 staining appeared to be most strongly counteracted when Hrs and Tsg101 were simultaneously depleted (Fig. 4B). Similarly to what was observed when Hrs or Tsg101 were singly depleted, under these conditions Cx43 was often found to localize in EEA1-positive endosomes (Fig. 5).

In agreement with previous studies, depletion of Hrs and/or Tsg101 resulted in larger endosomes (Fig. 5; and data not shown) (Razi and Futter, 2006). The degree of enlargement was greatest when both Hrs and Tsg101 were depleted. Importantly, under these conditions most Cx43 was found to localize at the rim of these enlarged endosomes (Fig. 5).

To gain a clearer understanding of the role of Hrs and Tsg101 in degradation of Cx43 gap junctions, the ultrastructural localization of Cx43 was determined by immunoelectron microscopy. In these experiments, control siRNA-transfected cells or cells depleted of Hrs and Tsg101 were treated with TPA and cycloheximide for 1.5 hours. Cells were then prepared for electron microscopy, and Cx43 was detected using immunogold particles. As expected, control cells were nearly completely devoid of Cx43 labelling under these conditions (data not shown). By contrast, cells depleted of Hrs and Tsg101 showed strong subcellular Cx43 labelling. In accordance with the confocal microscopy studies, Cx43 was frequently localized in endosomes (Fig. 6A; supplementary material Fig. S2A). In addition, Cx43 was often found to be organized as annular gap junctions (Fig. 6B; supplementary material Fig. S2B). Usually, the double membrane structure of these annular gap junctions appeared to be partly or completely disrupted, forming single membranes as well as small intralumimal vesicles. Importantly, such annular gap junctions often appeared to be in the process of fusing with other types of vesicles (Fig. 6C; supplementary material Fig. S2C). Other vesicles appeared to contain Cx43-enriched multivesicular structures, presumably remnants of annular gap junctions (Fig. 6D; supplementary material Fig. S2D).

Interestingly, after prolonged treatment with TPA, most Cx43 was found at the plasma membrane rather than in early endosomes (3 hours, Fig. 4B; 4.5 hours, supplementary material Fig. S3) in cells depleted of Hrs or Tsg101. Under these conditions Cx43 was organized as distinct gap junctions. Also when Hrs and Tsg101 were simultaneously depleted, a subpopulation of Cx43 was found to be organized as gap junctions after 3 and 4.5 hours treatment of TPA (Fig. 4B and supplementary material Fig. S3, respectively). However, most Cx43 was localized in intracellular vesicles under these conditions, in contrast to what was observed when Hrs and Tsg101 were singly depleted.

To substantiate the role of Hrs and Tsg101 in degradation of Cx43, we next examined how depletion of Hrs and/or Tsg101 affected the Cx43 protein level, as determined by western blotting. Depletion of Tsg101 caused an approximate 50% increase in the total Cx43 protein compared with control cells. By contrast, depletion of Hrs alone or in combination with Tsg101 did not significantly affect the level of Cx43 protein (Fig. 7).

We next evaluated how depletion of Hrs and Tsg101 affected the PKC-induced degradation of Cx43. As expected, degradation of Cx43 in control siRNA-transfected cells was strongly induced in response to TPA treatment. After 1.5 hours chase, the Cx43 protein level was approximately 25% that of the untreated cells (Fig. 7). Importantly, cells depleted of either Hrs or Tsg101 showed reduced degradation of Cx43 in response to TPA exposure, and were found to have a Cx43 protein level at this time point of approximately 60% and 80%, respectively, compared to untreated control cells. The strongest effect on TPA-induced degradation of Cx43 was observed when Hrs and Tsg101 were simultaneously depleted, in which the Cx43 protein level was approximately 90% compared with untreated control cells.

After 3 hours of TPA treatment, the Cx43 protein level in control siRNA-transfected cells had decreased to approximately 10% of that in the untreated control cells (Fig. 7). At this time point, the Cx43 protein level in cells depleted of Hrs or Tsg101 was approximately 40% and 55%, respectively, of that in the untreated control cells. Under conditions where Hrs and Tsg101 were simultaneously depleted, the Cx43 protein level was approximately 80%. Collectively, these data strongly suggest that Hrs and Tsg101 are both required for PKC-induced degradation of Cx43.

Interestingly, although both Hrs and Tsg101 depletion counteracted the PKC-induced degradation of Cx43, the SDS-PAGE band pattern of Cx43 differed under these conditions. Following 3 hours of TPA treatment, Cx43 was mainly found in the P0 and P1 states in Hrs-depleted cells, whereas in Tsg101-depleted cells, most Cx43 was in the P1 and P2 states (Fig. 7). When both Hrs and Tsg101 were depleted, Cx43 was predominantly found in the P2 state.

To further elucidate the role of Hrs and Tsg101 in the endocytic trafficking of Cx43, we examined how depletion of Hrs and Tsg101 affected the detergent resistance of Cx43. As expected, in untreated cells, most P1 and P2 forms of Cx43 were Triton X-100 insoluble (Fig. 8A). After 1.5 hours of treatment with TPA and cycloheximide, Cx43 was mostly found in a Triton X-100 soluble form, in accordance with the above data showing that at this time point Cx43 is mostly localized in intracellular vesicles (Figs 4 and 5). As expected, the TPA-induced solubility of Cx43 in Triton X-100 was not counteracted by Hrs and/or Tsg101 depletion. Interestingly, after 3 hours of treatment with TPA and cycloheximide, most Cx43 was again in a Triton X-100 resistant form in cells depleted of either Hrs or Tsg101. This observation is in accordance with the immunofluorescence data, showing that most Cx43 at this time point is reorganized as gap junctions at the plasma membrane (Fig. 4B). By contrast, when Hrs and Tsg101 were simultaneously depleted, a significant subpopulation of Cx43 remained in a Triton X-100 soluble state at this time point, in agreement with the data showing that under these conditions Cx43 is localized both intracellularly and in gap junctions (Fig. 4).

We next investigated whether the gap junctions present during prolonged TPA treatment in Hrs- or Tsg101-depleted cells are functional. Depletion of Hrs and Tsg101 alone or in combination did not have major effects on gap junctional communication under constituitive conditions (Fig. 8B). In accordance with previous studies, TPA nearly completely blocked gap junctional communication (supplementary material Fig. S4) (Rivedal and Opsahl, 2001). As expected, depletion of Hrs or Tsg101 did not affect the initial TPA-induced block in communication. Following 3 hours of exposure to TPA and cycloheximide, the gap junctional communication in control siRNA-transfected cells remained strongly downregulated (Fig. 8C). Importantly, cells depleted of Hrs or Tsg101 showed an approximate threefold increase in gap junctional communication compared with control siRNA-transfected cells at this time point. Taken together, these data suggest that the gap junctions present during prolonged TPA treatment in cells depleted of Hrs or Tsg101 are functional.

Cx43 ubiquitylation in response to TPA treatment has previously been shown to be transient (Leithe and Rivedal, 2004b). The mechanisms involved in the deubiquitylation of Cx43 are, however, not known. Since the above findings suggest an important role of Hrs and Tsg101 in the endocytic trafficking of Cx43, we asked whether these proteins could play a role in regulating the deubiquitylation of Cx43. Interestingly, under conditions where Hrs and Tsg101 were simultaneously depleted, Cx43 remained strongly ubiquitylated compared with control cells or cells singly depleted of Hrs or Tsg101 (Fig. 9A,B). This effect was observed both in untreated cells and cells treated with TPA for 1.5 or 3 hours. Importantly, Cx43 was found, under these conditions, to partly colocalize with ubiquitin at the rim of enlarged endosomes, as determined by confocal microscopy (Fig. 9C). These data suggest that either Hrs or Tsg101 must be present for Cx43 to undergo deubiquitylation. The data further suggest that under conditions where both Hrs and Tsg101 are absent, a subpopulation of ubiquitylated Cx43 accumulates at the rim of early endosomes.

Previous studies indicate an important role of the proteasome in trafficking of growth factor receptors from early endosomes to lysosomes (Bishop et al., 2002; Longva et al., 2002; Rocca et al., 2001). Since the above data suggest that early endosomes have a crucial role in trafficking of ubiquitylated Cx43 to lysosomes, we considered it important to elucidate the role of the proteasome in this process. We first investigated how proteasomal inhibitors affected the colocalization of ubiquitin and Cx43 gap junction plaques, since our previous data indicated that proteasomal inhibitors counteract TPA-induced ubiquitylation of Cx43 (Leithe and Rivedal, 2004b). Interestingly, the proteasomal inhibitor MG132 strongly counteracted ubiquitin recruitment to gap junctions, as determined by confocal microscopy (Fig. 10A). Proteasomal inhibitors have been reported to reduce the cellular level of free ubiquitin, because of prevention of deubiquitylation of proteasomal substrates (Schubert et al., 2000). In accordance with these observations, treating IAR20 cells with TPA in combination with MG132 for 15 minutes resulted in a strong reduction in the level of free ubiquitin (Fig. 10B). This observation suggests that the reduced recruitment of ubiquitin to gap junction plaques caused by proteasomal inhibition is due to depletion of the pool of free ubiquitin.

To investigate whether the proteasome is involved in trafficking of Cx43 from early endosomes to lysosomes, cells were treated with TPA for 30 minutes and subsequently incubated for 60 minutes with TPA alone or in combination with MG132 (Fig. 10B). As expected, in cells treated with only TPA for 90 minutes, most Cx43 staining was lost. Interestingly, this loss of Cx43 was not affected when TPA was co-incubated with MG132 for 60 minutes following the initial TPA treatment of 30 minutes. Collectively, these results suggest that once Cx43 has been ubiquitylated at the plasma membrane, proteasomal activity is not required for its trafficking from the early endosome to the lysosome.

Modulation of gap junction degradation is considered to be an important mechanism by which the level of intercellular communication via gap junctions is regulated (Berthoud et al., 2004; Laird, 2005; Musil et al., 2000; Thomas et al., 2003). However, the molecular machinery involved in the degradation of gap junctions has remained poorly understood. It was previously reported that EGF- and TPA-induced degradation of Cx43 is preceded by Cx43 ubiquitylation (Leithe and Rivedal, 2004a; Leithe and Rivedal, 2004b). In the present work, we have identified the ubiquitin-binding proteins Hrs and Tsg101 as crucial mediators of Cx43 trafficking to the lysosome. To the best of our knowledge, this study represents the first evidence that the early endosome functions as a sorting organelle for ubiquitylated Cx43. The data provide new insight into how cell-cell communication via gap junctions can be regulated at the level of connexin turnover.

Activation of PKC was found to be associated with a dramatic relocalization of ubiquitin to gap junction plaques, as determined by confocal microscopy. Ubiquitin recruitment was particularly prominent in gap junction regions that were in the process of internalizing into one of the adjacent cells. Often, entire gap junction plaques were coated with ubiquitin in response to PKC activation. Furthermore, internalization of gap junctions was found to occur not only in the centre of the plaque, but also in its periphery. Notably, this finding is in contrast to a previous study suggesting that only the centre of the plaque is internalized, whereas the more peripheral parts of the plaque do not undergo internalization (Gaietta et al., 2002). The discrepancies between the two studies may possibly reflect two different pathways for internalization of Cx43 gap junctions. The first pathway, in which only Cx43 in the centre of the plaque is internalized, might occur under constitutive conditions, possibly in a ubiquitin-independent manner. The second pathway, as reported here, might possibly be induced only in response to Cx43 phosphorylation and ubiquitylation. The finding that ubiquitin can be present in gap junction plaques is in agreement with a previous immunoelectron microscopy study by Rütz and Hülser (Rütz and Hülser, 2001).

Cx43 was found to remain ubiquitylated during its trafficking to early endosomes, and partly colocalized with Hrs and Tsg101. Depletion of Tsg101 caused increased levels of Cx43 protein under constitutive conditions. Interestingly, under these conditions Cx43 was not found to localize in endosomal compartments, but was instead organized as intercellular gap junctions. Furthermore, although depletion of Tsg101 counteracted TPA-induced trafficking of Cx43 from the early endosome, Cx43 appeared to be able to escape early endosomes and relocate to the plasma membrane during prolonged TPA treatment. Importantly, the localization of Cx43 at the plasma membrane during prolonged treatment with TPA was associated with a normalization of the Cx43 phosphorylation and ubiquitylation status. Furthermore, both the analysis of the detergent resistance of Cx43 and the dye transfer experiments indicate that the Cx43 population that escapes degradation under conditions where Tsg101 is depleted is able to form functional gap junctions. Further studies are required to elucidate the molecular mechanisms involved in the ability of Cx43 to form functional gap junctions during prolonged TPA treatment of cells depleted of Tsg101. However, one possibility could be that Cx43 undergoes recycling from early endosomes to the plasma membrane. A scenario in which Cx43 undergoes trafficking from the early endosome to the plasma membrane in cells lacking Tsg101 would be in accordance with what has been observed for other transmembrane proteins in other cell types. For instance, endocytosed EGF receptors that were normally sorted to the lysosome were instead rapidly recycled back to the cell surface in a Tsg101 mutant cell line (Babst et al., 2000). Enhanced recycling of EGF receptor was also observed when Tsg101 was depleted by RNA interference (Bishop et al., 2002; Raiborg et al., 2008). The most common method used for studying recycling of plasma membrane proteins is to covalently link biotin moieties to their extracellular domains followed by pulse-chase experiments. Biotinylation assays have been successfully used to show that Cx43 hemichannels undergo recycling under specific conditions (VanSlyke and Musil, 2005). However, the cellular compartment(s) in which recycling of Cx43 hemichannels occurs have not been identified. Unfortunately, biotinylation assays cannot be used to investigate endocytosis and recycling of gap junction-associated Cx43, since the assembly of connexin proteins into gap junction plaques prevents the binding of biotin to their extracellular domains. Thus, although the data in the present study opens the possibility that Cx43 can undergo recycling from the early endosomes to the plasma membrane, further studies are required to test this hypothesis.

While the present work was in progress, it was reported that Tsg101 binds to Cx31, Cx43 and Cx45 (Auth et al., 2009). It was also suggested that Tsg101 is involved in constitutive degradation of Cx43 and Cx45, but not Cx31. Our results are in accordance with the notion that Tsg101 plays a key role in Cx43 degradation under constitutive conditions.

In contrast to Tsg101, Hrs depletion did not counteract the constitutive degradation of Cx43. However, similarly to what was observed when Tsg101 was depleted, TPA-induced degradation of Cx43 was reduced in Hrs-depleted cells. Also under these conditions, most Cx43 was found to localize at the plasma membrane during prolonged TPA treatment. The TPA-induced degradation of Cx43 was most strongly counteracted when Hrs and Tsg101 were simultaneously depleted. Interestingly, under these conditions, most Cx43 appeared to remain localized in early endosomes in a phosphorylated and ubiquitylated state, even after prolonged TPA treatment. It should be noted that the phosphorylation events involved in the formation of the Cx43-P2 form are currently incompletely understood (Solan and Lampe, 2009). Thus, the Cx43 phosphorylation sites that are responsible for the formation of the Cx43-P2 band observed during prolonged TPA treatment of cells depleted of Hrs and Tsg101 remain to be determined. Further studies are also required to elucidate the relationship between Cx43 phosphorylation and ubiquitylation, both under constitutive conditions and in response to TPA treatment.

Taken together, the present data identify the early endosome as a major sorting station for ubiquitylated Cx43 and reveal a crucial role of Hrs and Tsg101 in trafficking of Cx43 to lysosomes. It is thought that Hrs and Tsg101 mediate trafficking of growth factor receptors by binding directly to ubiquitylated receptors in the early endosome (Raiborg and Stenmark, 2009). In analogy to their role in downregulation of growth factor receptors, we propose a model in which Hrs and Tsg101 mediate lysosomal degradation of Cx43 by binding to ubiquitylated Cx43 (Fig. 11). This study further suggests that either Hrs or Tsg101 must be present for Cx43 to undergo deubiquitylation, possibly by recruiting enzymes catalyzing Cx43 deubiquitylation.

The human genome encodes 84 predicted active deubiquitylation enzymes, several of which have been shown to localize at early endosomes and mediate deubiquitylation of internalized receptors (Saksena et al., 2007). An important subject for future research will be to identify the enzymes catalyzing Cx43 deubiquitylation, and their potential role in regulating the level of gap junctional communication.

The rat liver epithelial cell line IAR20, originally isolated from normal inbred BD-IV rats (Asamoto et al., 1991; Montesano et al., 1977), was obtained from the International Agency for Research on Cancer (Lyon, France). Cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% (v/v) fetal bovine serum (FBS; Gibco BRL Life Technologies, Inchinnan, UK). The growth medium was replaced with DMEM supplemented with 1% (v/v) FBS 24 hours prior to experiments.

The siRNA oligonucleotides targeted against Hrs or Tsg101 were obtained from Invitrogen (Stealth Select RNAi HGSRSS329127 and TSG101RSS333993, respectively). The Hrs siRNA had the following sequence: 5′-UUAUCAUUGACCUUCUUCUUGAUGG-3′. The sequence of the Tsg101 siRNA was: 5′-CCAGGCAGAGCUUAAUGCCUUGAAA-3′. The siRNA control construct was Stealth RNAi Negative Control Medium GC (Invitrogen). siRNA was transfected into cells using Lipofectamine 2000 (Invitrogen), according to the manufacturer's direction, at a final concentration of 80 nM. In some experiments, siRNA against Hrs and Tsg101 were co-transfected into cells, at a total final concentration of 80 nM. IAR20 cells grown in 35 mm Petri dishes were transfected 24 hours after seeding. The growth medium was replaced with DMEM supplemented with 1% (v/v) FBS 24 hours after transfection. The cells were assayed 48 hours after transfection. To ensure that the effect of Hrs and Tsg101 depletion on Cx43 as reported in the present study was not a result of an off-target effect, we repeated the experiments with two independent siRNA duplexes. In these experiments, we obtained a similar effect on Hrs and Tsg101 depletion and on Cx43 trafficking as with the siRNA duplexes described above, indicating that the observed effect of siRNA on Cx43 was indeed due to depletion of Hrs and Tsg101 (data not shown).

TPA, cycloheximide, Lucifer yellow and chlordane were obtained from Sigma (St Louis, MO). GF109203 was from Calbiochem (La Jolla, CA). Anti-Cx43 antibodies were obtained by injecting rabbits with a synthetic peptide consisting of the 20 C-terminal amino acids of Cx43 (Rivedal et al., 1996). Mouse anti-Cx43 antibodies were from Chemicon International (Temecula, CA). The P4D1 (mouse IgG) and FK2 (mouse IgG) anti-ubiquitin antibodies were obtained from Covance (Berkeley, CA) and Affinity Research Products (Exeter, UK), respectively. Mouse anti-early endosomal autoantigen-1 (EEA1) antibodies were from BD transduction laboratories (San Diego, CA). Rabbit anti-Hrs antibodies have been described previously (Raiborg et al., 2001). Mouse anti-Tsg101 antibodies were purchased from Genetex (San Antonio, TX) or Abcam (4A10, Cambridge, UK). Mouse anti-actin antibodies were from Sigma. Alexa-Fluor-488-conjugated goat anti-rabbit IgG and Alexa-Fluor-594-conjugated goat anti-mouse IgG were from Molecular probes (Eugene, OR). Horseradish peroxidase-conjugated goat anti-rabbit IgG secondary antibodies were from Bio-Rad (Hercules, CA). Horseradish peroxidase-conjugated donkey anti-mouse IgG antibodies were obtained from Jackson Immunoresearch Laboratories (West Grove, PA).

Western blotting and immunoprecipitation was performed as described previously (Leithe and Rivedal, 2004a). The intensities of Cx43 bands were quantified using Scion Image (Scion). In some experiments, cells were subjected to a Triton X-100 solubility assay, as described below, before immunoprecipitation.

The Triton X-100 solubility of Cx43 was determined based on a method described by VanSlyke and Musil (VanSlyke and Musil, 2000), as described previously (Sirnes et al., 2008). Briefly, cells were collected in 1 ml incubation buffer (136.8 mM NaCl, 5.4 mM KCl, 0.34 mM Na₂HPO₄, 0.35 mM KH₂PO₄, 0.8 mM MgSO₄, 2.7 mM CaCl₂, 20 mM HEPES, pH 7.4) containing 10 mM N-ethylmaleimide and 200 μM phenylmethylsulfonyl fluoride and centrifuged at 3000 g for 4 minutes at 4°C. The cells were resuspended in 400 μl incubation buffer containing 10 mM N-ethylmaleimide, 200 μM phenylmethylsulfonyl fluoride, 22 μM leupeptin and phosphatase cocktail, and sonicated for 10 seconds. Triton X-100 was added to a final concentration of 1% (v/v). Lysates were incubated for 30 minutes at 4°C. A 200 μl aliquot of the sample was then transferred to a microcentrifuge tube and centrifuged at 100,000 g for 50 minutes at 4°C. The supernatant fraction was collected and the pellet was resuspended in 200 μl incubation buffer containing 10 mM N-ethylmaleimide and 200 μM phenylmethylsulfonyl fluoride. To all three fractions (total cell lysate, Triton-X-100-soluble and Triton-X-100-insoluble), 200 μl sample buffer containing 2× SDS was added and heated for 5 minutes at 95°C. Western blot analysis was performed as described above.

Cells were fixed and stained as described previously (Leithe and Rivedal, 2004a). Nuclei were stained with Hoechst 33342. Immunofluorescence images were captured using a Nikon E800 microscope with a Spot-2 camera. For colocalization studies, cells fixed on coverslips were analyzed with a LSM 510 META confocal microscope (Carl Zeiss) equipped with Plan Apochromat 63× 1.4 NA and Neo Fluar 100× 1.45 NA oil immersion objectives (Carl Zeiss). Appropriate emission filter settings were included to exclude bleed-through effects. Images were acquired with the LSM 510 software (version 3.2; Carl Zeiss) and processed with Adobe Photoshop, version 7.0.

Cells were prepared for immunoelectron microscopy as described previously (Leithe et al., 2006a). Cryosections were transferred to formvar-carbon-coated grids and labelled with anti-Cx43 antibodies followed by protein A-gold conjugates. The labelled cryosections were examined with a Philips CM10 or a JEM 1230 (JEOL) electron microscope.

Quantitative scrape loading was performed as previously described (Leithe et al., 2003; Opsahl and Rivedal, 2000). Digital monochrome images were acquired using a COHU 4912 CCD camera (COHU, San Diego, CA) and a Scion LG-3 frame grabber card (Scion Corporation, Frederick, MD).