Helicobacter pylori VacA induces apoptosis by accumulation of connexin 43 in autophagic vesicles via a Rac1/ERK-dependent pathway

Helicobacter pylori (H. pylori) produces vacuolating cytotoxin (VacA), a potent protein toxin, which is associated with gastric inflammation and ulceration. Recent studies demonstrated that connexins (Cxs), which are responsible for intracellular communication at gap junctions (GJs) as well as cell homeostasis, participate in VacA-induced cell death. We now demonstrate in AZ-521 cells that VacA increased cytoplasmic Cx43, accompanied by LC3-II generation in a time- and dose-dependent manner without induction of Cx43 mRNA expression. Inhibition of VacA-induced Rac1 activity prevented ERK phosphorylation and the increase in Cx43. Suppression of ERK activity and addition of N-acetyl-cysteine inhibited VacA-dependent increase in Cx43 and LC3-II. DIDS, an anion-selective inhibitor, suppressed VacA-dependent increase in Cx43, suggesting that VacA channel activity was involved in this pathway. By confocal microscopy, Cx43 increased by VacA was predominately localized in cholesterol-rich, detergent-resistant membranes including GJs, and a fraction of Cx43 was incorporated in endocytotic vesicles and autophagolysosomes. Accumulation of Cx43 was also observed in gastric mucosa from H. pylori-infected patients compared with healthy controls, suggesting that the pathogen caused a similar effect in vivo. Our findings show that VacA-mediated effects on autophagy inhibits turnover of Cx43, resulting in increased levels in the cytoplasm, leading eventually to apoptotic cell death.


INTRODUCTION
Helicobacter pylori (H. pylori) is a gram-negative bacterium that causes persistent infection in the gastric mucosa, with associated inflammation, gastric and duodenal ulceration, MALT lymphoma, and gastric cancer. 1 H. pylori produces a number of virulence factors that are involved in pathogenesis of disease. 2,3 Among them, vacuolating cytotoxin, VacA, is an exotoxin of about 90 kDa, which has been shown to be associated with gastric inflammation and ulceration in animal models of disease. 4 VacA has two functional domains (p33-37 and p55-58). The p33-37 domain contributes to cytotoxicity and the p55-58 domain is responsible for binding to target cell receptors, for example, sphingomyelin, fibronectin, receptor protein-tyrosine phosphatase α (RPTPα), RPTPβ, low-density lipoprotein receptor-related protein 1 (LRP1). 3,5 Numerous studies have shown that VacA disrupts cell functions (e.g., autophagy, channel formation, intracellular vesicle traffic antigen presentation), leading eventually to cell death. 3,6 Connexin 43 (Cx43), a member of the human Cx family, which is ubiquitously expressed and the most intensively studied of the Cx family, 7 is responsible for VacA-induced cell death; its protein level is associated with sensitivity to VacA. 8 In general, Cxs are integral membrane proteins, which form channels at gap junctions (GJs). 9,10 They regulate intercellular communication, cell-cell channel formation and exchange of signaling molecules. This intercellular communication has a critical role in development and homeostasis. 10,11 The amino-acid sequence and length of the C-terminal domain of Cx43 is unique compared with other Cxs. 12 This C-terminal region has multiple modification sites (e.g., phosphorylation, ubiquitination, SUMOylation) 7 and contributes to interaction with a number of functional proteins in GJs during pH-dependent channel closure. 13,14 In contrast, the N-terminal domains of Cxs are involved in transjunctional voltage-dependent gating of GJ channels 15 and oligomerization. 16 Mutations in the N-terminal domain of Cx43 inhibit functional channel formation, and are found in a human developmental disorder known as oculodentodigital dysplasia. 14 Furthermore, abnormal upregulation of Cx43 has been observed in diseased tissues (e.g., diabetic skin, 17 blood vessels in proximity to a site of injury, 18 arthritis, 19 breast cancer and melanoma. 20 Cx43-mediated cell-cell communication in GJs has an essential role in peritoneal metastasis. 21 Interestingly, Cxs turnover is faster than that of the other transmembrane proteins, for example, the half-life of Cx43 in cultured cells is 1-5 h and is regulated by pathways involving proteasomes, lysosomes and autophagy. [22][23][24] Autophagy is an important intracellular degradation system for maintaining cellular homeostasis. 25 VacA-induced autophagy requires its channel-forming activity and induction of reactive oxygen species (ROS). 26,27 We previously identified LRP1 as a receptor mediating VacA-induced autophagy, followed by apoptosis. 5 In addition, Tsugawa et al. reported that H. pylori CagA protein, which is a type IV secretion effector and associated with the development of gastric cancer, was decreased by VacA-induced autophagy. Binding of VacA to LRP1 in AGS cells resulted in production of ROS and induction of autophagy, leading to CagA degradation. 5,28 Interestingly, in some cells, Cx43 contributes to induction of autophagy 24,29 and apoptosis. 30 However, it is not clear whether Cx43 is associated with VacA-induced apoptosis and autophagy.
In the current study, we assessed the role of Cx43 in VacAinduced AZ-521 cell death and its presence in H. pylori-infected gastric mucosa. VacA did not affect Cx43 mRNA expression in AZ-521 cells, however, Cx43 accumulated in cytoplasm through slowed turnover by dysfunctional autophagy, leading eventually to apoptotic cell death in a glutathione (GSH)-and Rac1/ERKdependent and LRP1-independent manner. Finally, consistent with our findings in cultured cells, we demonstrate increased Cx43 in the gastric mucosa of H. pylori-infected patients.

RESULTS
VacA positively regulates Cx43 in AZ-521 cells Recently, Radin et al. 8 showed that the expression level of Cx43 in cells contributes to VacA-induced cell death. To confirm this finding, we investigated the effect of Cx43 knockdown on VacA-induced apoptosis (Figure 1a). After incubation with VacA for 8 h, we found increased Cx43, associated with cleavage of caspase-9 (cCas9) and PARP (cPARP). In contrast, enhanced caspase-9 and PARP cleavage by VacA was significantly suppressed in Cx43knockdown cells. In agreement, VacA induced Bax conformational changes in control cells, but not in Cx43-knockdown cells (Figure 1b). Previous studies reported that Cx43-mediated apoptosis is regulated by Bcl-2. 31,32 We found that the basal expression levels of apoptosis blocker, Bcl-2 and Bcl-xL, were significantly increased in Cx43-knockdown cells compared with control cells (Figure 1c). In addition, VacA-induced PARP cleavage was suppressed in FLAGtagged Bcl-xL-expressing cells, but did not affect VacA-dependent increase in Cx43 (Figure 1d). Induction of Cx43 in VacA-treated cells was observed after an 8 h incubation, whereas LC3-II generation, which is an index of autophagy and preceded apoptosis in VacAtreated cells, 5 occurred in a 4 h incubation (Figure 1e). Induction of both Cx43 and LC3-II in VacA-treated cells were concentration dependent. Cx43 induction was significant at VacA concentrations 45 μg/ml VacA ( Figure 1f). Next, to assess whether VacA enhanced Cx43 transcription, we determined Cx43mRNA production by realtime qPCR (Figure 1g). The levels of Cx43 mRNA were not altered by VacA after 8 h incubation, suggesting that VacA may increase Cx43 accumulation by inhibiting its degradation, rather than by affecting its translation.
Phosphorylation and ubiquitination of the increased cytoplasmic Cx43 observed in the presence of VacA It is well known that Cxs turnover is faster than that of other transmembrane proteins and its degradation is carried out by proteasome, endosome/lysosome and autophagy. [22][23][24] As phosphorylation of Cx43 facilitated its degradation by the proteasome, we investigated whether proteasome inhibition affected the VacAinduced Cx43 increase. AZ-521 cells were treated with the proteasome inhibitors, MG132 and lactacystin, for 30 min followed by co-incubation with or without VacA (Figure 2a). VacA increased the Cx43-P 0 band (non-phosphorylated Cx43) in control cells, whereas the Cx43-P X band (phosphorylated Cx43) was detectable only when cells were treated with proteasome inhibitors. Proteasome inhibition blocked the degradation of Cx43-P X in control cells. In VacA-treated cells, proteasome inhibitors did not block the degradation of Cx43-P X and Cx43-P 0 maintained its basal level in the presence of proteasome inhibitors. Thus, VacA did not stimulate the phosphorylation of Cx43, which is involved in regulation of GJ turnover and function. Proteasome inhibitors did not alter VacA-induced cPARP and cCas9, suggesting that phosphorylated Cx43 does not affect VacA-induced apoptosis. We next investigated if Cx43 was modified by ubiquitin. After cells were incubated with or without VacA, cell lysates were immunoprecipitated with anti-Cx43 antibody, and then analyzed with anti-ubiquitin antibodies (FK1 and P4D1). The P4D1 antibody recognizes all forms of ubiquitin, whereas FK1 antibody recognizes polyubiquitin chains. 33 As shown in Figure 2b, VacA increased ubiquitinated Cx43, which is sorted through the endosomal or autophagic pathway, moving then to lysosomes where Cx43 is degraded. Our findings suggested that proteasomes did not affect autophagy and apoptosis via Cx43 in VacA-treated cells.
Localization of VacA-increased Cx43 in autophagic vesicles, but not in mitochondria By confocal microscopy, Cx43 localized at GJs in plasma membranes of control cells, while Cx43 accumulated in cytoplasmic compartments and colocalized with autophagosomal marker LC3 in VacA-treated cells (Figure 3a), suggesting that VacAinduced autophagy may be critical for apoptosis via Cx43. We also observed many puncta of LC3 and Cx43 and LysoTracker-positive vesicles in VacA-treated cells. All Cx43 colocalized with LC3 (white arrows), while some puncta of Cx43 with LC3 were not observed in autophagolysosome stained with LysoTracker, as shown by yellow arrows (Figure 3b). Silencing of the Cx43 gene with siRNA resulted, as expected, in interference with VacA-increased Cx43. On the basis of analysis of LC3-II generation by western blotting, however, generation of LC3-II and LysoTracker-positive vesicles was not inhibited by Cx43 knockdown, (Figure 3c), suggesting that there is another pathway that mediates VacA-induced autophagic vesicles independent of Cx43. VacA is degraded by autophagy after binding to LRP1, but not RPTPs, on the cell surface. 5 Consistent with these findings, inhibition of LRP1 expression with LRP1 siRNA did not affect VacA-induced increase in cytoplasmic Cx43. Treatment of starved cells with an autophagy inhibitor, such as 3-methyladenine or chloroquine, increased the amount of Cx43 in GJs, 34 which was resistant to extraction with Tx (Tx-insoluble fraction), indicating that blocking autophagy specifically prevented the loss of Cx43 at the plasma membrane. Cx43 was detected in both Tx-insoluble and Tx-soluble fractions of control cells, whereas VacA-increased Cx43 was enriched in Tx-insoluble fraction ( Figure 3d). In agreement, confocal microscopic analysis showed that both VacA and VacA-increased Cx43 in cytoplasm were resistant to Tx extraction, whereas only Cx43 in GJs of control cells was resistant to this detergent (Figure 3e). By monitoring Alexa488-labeled transferrin, which is internalized by dynamindependent endocytosis, 35 we found that Alexa488-labeled transferrin was completely extracted by Tx. In addition, we assessed whether VacA-induced Cx43 was colocalized with LC3 and VacA in these Tx-extraction-resistant compartments. As shown in Figure 3f, Cx43 colocalized with both LC3 and VacA in control cells. After treatment with T × 100 buffer, most LC3 was lost, however, the vesicles containing both VacA and Cx43 were resistant to T × 100 extraction. These data suggest that Cx43 and VacA were in vesicles containing cholesterol-rich membranes, which are resistant to Tx extraction. 36 Taken together, Cx43, increased by VacA, localized in cytoplasmic vesicles, which consist of cholesterol-rich membranes as was found in GJs. Cx43, which localized in the mitochondria, influenced mitochondrial integrity and initiation of apoptosis. 37,38 Cx43 also colocalized with autophagosome markers such as LC3 39 and ATG16L1. 27 By confocal microscopy, Cx43 increased by VacA colocalized with EEA1, LAMP1, Atg16L1 and LC3, suggesting that Cx43 is associated with cellular trafficking pathways involving endosomes and autophagy. However, VacA-increased Cx43 did not localize in mitochondria, which were detected by MitoTracker staining (Figures 3g-j).
Atg16L1 and ROS/ERK signaling pathway participates in VacAinduced Cx43 increase We next examined whether autophagy regulates VacA-dependent increase in Cx43. VacA induced LC3-II generation independent of the increase in Cx43 (Figures 3b and c). Silencing of the Atg16L1 gene with Atg16L1 siRNA resulted in suppression of both Cx43 increase and LC3-II generation in VacA-treated cells as compared with control siRNA-transfected cells (Figure 4a). Thus, our results indicate that Atg16L1, which has an essential role in autophagy, 40 participated in VacA-induced Cx43 increase and LC3-II generation. In addition, depletion of Atg16L1 significantly suppressed VacAinduced PARP cleavage (Figure 4b).    ROS is a key factor in induction of autophagy 41 and the increase of Cx43 in cultured cardiomyocytes. 42 A prior study demonstrated that VacA-reduced GSH is a trigger to induce LC3-II generation in AGS cells. 28 N-acetyl-cysteine (NAC) is an antioxidant and free radical scavenger that increases intracellular GSH. 43 We assessed whether GSH regulates VacA-induced Cx43 increase in AZ-521 cells and found that VacA-induced Cx43 increase and LC3-II generation were significantly suppressed by incubation with 10 mM NAC ( Figure 4c) and 10 mM GSH ( Figure 4d).
MAPK also regulates autophagy. 44 Previous studies showed that VacA activates the ERK/MAP kinase cascade. 45 Next, we investigated the effect of NAC on VacA-induced ERK activity. In control cells, VacA transiently increased ERK phosphorylation after 0.5 h incubation, with a return to basal level after 2 h. VacA-induced ERK phosphorylation, however, was suppressed in NAC-treated cells ( Figure 4e). Furthermore, cells incubated with the ERK inhibitor, U0126 and VacA showed a significant decrease in the levels of Cx43 and LC3-II compared with cells treated with VacA only (Figure 4f). The ERK inhibitor did not affect VacA-induced vacuolating activity (Supplementary Figure S1b). Furthermore, ERK knockdown inhibited VacA-induced Cx43 increase and LC3-II generation ( Figure 4g). ERK knockdown, however, did not affect Cx43 mRNA level (Figure 4h). We also monitored by confocal microscopy the effect of ERK knockdown on VacA-increased Cx43. VacA increased accumulation of Cx43 in control cells, whereas VacA did not increase accumulation of Cx43 puncta in ERKknockdown cells (Figure 4i). More interestingly, ERK knockdown significantly suppressed VacA-induced cleavages of PARP and caspase-9 ( Figure 4j). These data indicated that GSH level controls ERK activation, and thereby regulates VacA-increased Cx43 and LC3-II generation as well as apoptosis.
Rac1 signaling pathway regulates VacA-induced Cx43 increase Hotchin et al. 46 showed that VacA activity is regulated by Rac1, a small GTP-binding protein that controls actin cytoskeleton reorganization and signal transduction. Furthermore, cytoskeletal reorganization is regulated by ROS, important mediators acting downstream of Rac1. 47 We investigated here whether Rac1 is involved in the VacA-induced Cx43 increase. AZ-521 cells treated with the Rac1 inhibitor, NSC23766, in the presence of VacA, showed a significant decrease in the levels of Cx43 compared with treatment with VacA alone (Figure 5a). Furthermore, Rac1 knockdown inhibited VacA-induced Cx43 increase ( Figure 5b). We further examined the effect of Rac1 on VacA-induced ERK phosphorylation. In the presence of the Rac1 inhibitor or following Rac1 knockdown, VacA-induced ERK phosphorylation was suppressed compared with treatment with VacA alone (Figures 5c and d). We next investigated whether VacA induced Rac1 activation. AZ-521 cells were incubated with VacA in the presence or absence of NAC. In control cells, VacA enhanced formation of Rac1-GTP in a time-dependent manner, which was suppressed in the presence of NAC (Figure 5e).
Effects of anion-channel blocker and LRP1 knockdown on VacAinduced Cx43 increase Chloride channel blocker, DIDS, inhibited VacA-mediated channel activity and cellular vacuolation 48,49 as well as LC3-II generation. 5 We investigated the effect of 4,4′-disothiocyanatostibene-2,2′-  Figure 6a). In addition, VacA-induced ERK phosphorylation was also inhibited by DIDS (Figure 6b), suggesting that VacA-mediated channel activity triggered phosphorylation of ERK. VacA bound to specific receptors, 3,5 which are involved in several pathological events. We investigated if LRP1 was the receptor associated with VacA-induced ERK phosphorylation. ERK phosphorylation by VacA was not affected by LRP1 knockdown (Figure 6c). As shown in Figure 4f, Cx43 and Cx43 puncta/30 cells Experiments were repeated three times with similar results. (e) AZ-521 cells were incubated with 120 nM heat-inactivated (iV) or wild-type VacA (V) for 10 h and were treated with or without cold 1% Tx, fixed with 4% PFA, and reacted with anti-Cx43 antibodies (red) and incubated with DAPI (cyan) as described in Materials and Methods. Alexa488-labeled Transferrin was used as a positive control. Bars represent 20 μm. Experiments were repeated three times with similar results. (f) AZ-521 cells were incubated with Alexa555-labeled VacA for 10 h and then were treated with or without cold 1% T × 100, fixed with 4% PFA, and reacted with anti-Cx43 antibodies (green), anti-LC3 antibodies (blue) and with incubated DAPI (cyan). Bars represent 20 μm. Experiments were repeated three times with similar results. The arrows indicate the colocalization of VacA, Cx43 and LC3. (g) AZ-521 cells were incubated with 120 nM heat-inactivated (iV) or wild-type VacA (V) for 10 h and then reacted with anti-EEA1 or anti-LAMP1 (red), or anti-Cx43 antibodies as indicated (green) and incubated with DAPI. Bars represent 20 μm. Experiments were repeated three times with similar results. The arrows indicate the colocalization of Cx43 and EEA1 or LAMP1. (h) AZ-521 cells were incubated with 120 nM heat-inactivated (iV) or wild-type VacA (V) for 10 h; 50 nM MitoTracker (red) was added to cells before fixation as described in Materials and Methods. Cells were reacted with anti-Cx43 antibodies (green) and incubated with DAPI (cyan). Merged and higher magnification images of the outlined areas are shown. Bars represent 20 μm. Experiments were repeated two times with similar results. (i) AZ-521 cells were incubated with 120 nM heat-inactivated (iV) or wild-type VacA (V) for 10 h, then reacted with anti-Atg16L1 antibodies (red), and anti-Cx43 antibodies (green) and incubated with DAPI. Bars represent 20 μm. Experiments were repeated three times with similar results. The arrows indicate the colocalization with Cx43 and Atg16L1. (j) AZ-521 cells were incubated with 120 nM heat-inactivated (iV) or wild-type VacA (V) for 10 h. Cells reacted with anti-clathrin antibodies (green) and incubated with DAPI (cyan). Bars represent 20 μm. Experiments were repeated two times with similar results.

VacA regulates Cx43 content K Yahiro et al
VacA were in vesicles containing detergent-resistant membranes (DRMs). We next investigated whether LRP1 and LC3-II also localized in DRMs. In control cells, VacA-induced LC3-II and internalized LRP1 were colocalized in cytoplasmic compartments as shown in a previous report. 5 In contrast, both LRP1 and LC3-II were completely extracted by Tx, suggesting that LC3-II and LRP1 were found in cytoplasmic vesicles that did not include cholesterol-rich membranes (Figure 6d).
Although LRP1 knockdown suppressed LC3-II generation, 5 VacA-induced Cx43 increase was not affected by LRP1 knockdown, indicating that LRP1 is not responsible for VacA-induced Cx43 increase (Figure 6e). Thus, ERK activation through LRP1 may be not associated with the signal transduction pathway responsible for VacA-induced Cx43 increase. Similarly, depletion of other VacA receptors such as RPTPα, RPTPβ and fibronectin did not affect VacA-induced Cx43 increase and LC3-II generation (Figures 6f and g). These results raise the possibility that there might be a yet-to-be defined VacA receptor, which is responsible for the Cx43 increase.
Increase of Cx43 in human biopsy samples in H. pylori-infected human gastric mucosa To assess the above findings in H. pylori-infected human mucosa, we investigated Cx43 protein level in the biopsy samples. The demographic characteristic of the patients is shown in Supplementary Table 1. Eleven H. pylori-positive and five H. pylori-negative biopsy samples from gastric antrum were examined. Interestingly, Cx43 content was not significantly increased in gastric mucosa of H. pylori-negative patients. In contrast, Cx43 was clearly observed in the gastric epithelium region in 8 out of 11 H. pylori-positive biopsy specimens (Figure 7), which was a statistically significant difference compared with H. pylori-negative mucosa (P = 0.0256, Fisher's exact test, H. pyloripositive versus -negative mucosa). These results suggested that Cx43 significantly accumulated in H. pylori-infected human gastric mucosa, with a potential role in the pathogenesis of human peptic diseases caused by H. pylori.

DISCUSSION
Cx43 is a ubiquitously expressed component of GJ channels and has an important role in intracellular communication involving cell death and cell survival. 10 Cx43 overexpression in HeLa cells resulted in increased susceptibility to cell death caused by VacA. In addition, Cx43 knockdown by shRNA increased cellular resistance to VacA-induced cell death. 8 Streptonigrin and α-Fas also enhanced cell death of Cx43-transfected HeLa cells compared with their wild-type counterparts. 30 Our data show that incubation of AZ-521 cells with VacA for 4-12 h resulted in increased accumulation of Cx43 in the cytoplasm in association with apoptosis and autophagy. The accumulation of Cx43 occurred, even though Cx43 is rapidly degraded with a half-life of only a few hours. 50 Consistent with a previous study, 8  VacA regulates Cx43 content K Yahiro et al by chemotherapeutic agents. 31 Indeed, we showed that VacAinduced PARP cleavage was significantly inhibited in Bcl-xLexpressing cells. Thus, our findings indicate that Cx43 is involved in VacA-induced apoptosis, which is regulated by Bcl-family proteins. The fact that we did not observe Cx43 in mitochondria stained with MitoTracker suggests that VacA did not increase mitochondrial Cx43, which is known to modulate apoptosis ( Figure 3h). Furthermore, VacA treatment did not affect Cx43 mRNA content. These results are consistent with our hypothesis that VacA modulates Cx43 turnover through effects on its degradation, with Cx43 accumulation leading eventually to apoptotic cell death. On the other hand, Cx43 has cytoprotective effects against cell death, for example, loss of Cx43 expression led to astrocytic 51 and osteocytic death. 52 These diverse findings are consistent with the hypothesis that Cx43 signaling cascade might have different effects in different cell types. Beside regulation through differential gene transcription, it has been shown that Cx43 protein levels can be altered through lysosome-and proteasome-dependent degradation pathways. Confocal microscopic analysis showed that VacA causes the internalization of Cx43 from the plasma membrane, with reduction of membrane-localized Cx43, followed by accumulation of Cx43 in the cytoplasmic compartment, with several vesicle marker proteins (e.g., LC3, Atg16L1, EEA1 and LAMP1; Figures 3g and i). Of note, Cx43 knockdown did not affect VacA-induced LC3-II generation, whereas knockdown of both endosomal and autophagic marker proteins resulted in suppression of VacAinduced Cx43 increase in endosomal and autophagic vesicles. These data are consistent with the proposal that Cx43 was at least in part initially translocated by endocytosis to early endosomes, followed by translocation into autophagic vesicles, and then to lysosomes. Taken together, these results lead us to hypothesize that Cx43 turnover may be regulated through an endosome/ autophagy pathway (Figure 8).
In addition, the VacA-induced increase in Cx43 was mainly localized in a Tx-insoluble fraction. Its cytoplasmic localization was in cholesterol-rich, DRMs, with a double-membrane structure as seen in GJ plaques 24 . Accumulation of Cx43 with VacA in DRMs seems to be due to the resistance to degradation by autophagy, followed by an increase in cytoplasmic Cx43, leading eventually to apoptosis. Indeed, a previous study showed that elevated Cx43 sensitizes cells to ER stress-induced, caspase-dependent apoptosis. 53 Clathrin mediates both steps of internalization and recycling of Cx43 between GJ and early endosomes. 10 Dynamin is essential for clathrin-dependent, coated-vesicle formation. 54 We found that cells treated with dynasore, a dynamin inhibitor, showed inhibition of VacA-induced Cx43 increase although VacA still induced LC3-II accumulation (Supplementary Figure S2). VacA also caused the extraordinary clumping of clathrin (Figure 3j). These findings indicate that accumulation of cytoplasmic Cx43 by VacA may occur from the plasma membrane by a dynamin-dependent pathway. Thus, inhibition of clathrin-mediated turnover of Cx43 by VacA may cause the remodeling of Cx43-containing GJs.
We observed that VacA increased non-phosphorylated Cx43-P 0 as shown in Figure 2a, consistent with a previous study, 55 which showed that inhibition of proteasomal function by MG132 or lactacystin led to an increase in phosphorylated Cx43-P x , whereas VacA decreased Cx43-P x . These discrepant results suggest that, under control conditions, phosphorylated Cx43 is primarily eliminated by proteasomes. Kraft et al. 56 reported that active crosstalk occurred between proteasome-mediated degradation and selective autophagy. However, in the presence of VacA, Cx43-P x may be degraded by VacA-activated autophagolysosomes.  Interestingly, VacA stabilized Cx43-P 0 even in autophagolysosomes stained with LysoTracker Red, indicating that, in the presence of VacA, Cx43-P 0 becomes resistant to autophagic degradation due to a change in its presence in cholesterol-rich membranes.
In cardiac myocytes, localization of Cx43 is regulated through the Rac1 pathway. 57 A recent study suggested that Mena, a member of the Ena/VASP family of actin regulatory proteins, is controlled by Cx43 localization via regulation of Rac1. 58 In addition, cytoskeleton reorganization is regulated by ROS, which act downstream of Rac1. 47 These findings led us to investigate the possibility that regulation of Cx43 occurs downstream of Rac1. After confirming the effect of Rac1 inhibitor on VacA-induced vacuole formation 46 (Supplementary Figure S1c), we showed that Rac1 inhibition and Rac1 knockdown significantly suppressed VacA-induced Cx43 increase (Figures 5a and b) and VacA-induced ERK phosphorylation (Figures 5c and d); these effects of VacA were not observed in Cdc42-and Rho-knockdown cells (Supplementary Figure S3). We next demonstrated that VacA-induced Rac1 activation was decreased in the presence of NAC (Figure 5e). These results indicate that VacA-induced Rac1 activation controls ERK activation, resulting in an increase in Cx43. GSH acts as an antioxidant by directly interacting with ROS. 59,60 Our previous study demonstrated that VacA suppressed the turnover rate of intracellular GSH, indicating that VacA impairs GSH metabolism in AZ-521 cells. 61 In this study, we showed that VacA-induced Cx43 increase was significantly decreased by addition of GSH and NAC, suggesting that, in VacA-treated cells, disordered GSH metabolism triggers a Cx43 increase. In addition, it has been reported that the MAPK pathway is activated by depletion of GSH, followed by ROS generation. 62,63 Furthermore, cultured cardiomyocytes treated with ROS showed an increase in Cx43 expression. 42 Interleukin-1β increased Cx43 expression in synovial fibroblasts via an ERK-dependent pathway. 64 In addition, a lipidsoluble pesticide, Lindane, activated ERK followed by induction of aberrant Cx43 endocytosis in 42GPA9 Sertoli cells. 65 Despite our previous finding that LRP1 mediates VacA-induced LC3-II increase, 5 LRP1 knockdown did not block VacA-induced ERK activation (Figure 6c), suggesting that there are at least two pathways, ERKdependent and ERK-independent, to induce LC3-II generation by VacA and that ERK activation through LRP1 may not be responsible for VacA-induced Cx43 increase (Figure 6e). Thus, these findings suggest that VacA-induced Cx43 increase and LC3-II generation are associated with a ROS-dependent ERK signaling cascade.
H. pylori infection has an important role in pathogenesis of not only stomach or duodenal 66 but also a variety of skin 67 and lung diseases. 68 Thus, it seems that H. pylori causes systemic disease. Abnormal upregulation of Cx43 has been observed in several diseases. [17][18][19][20][21] Interestingly, reduction of Cx43 expression has been shown to be associated with enhanced wound closure. [69][70][71] Our study demonstrated the elevated Cx43 in H. pylori-infected human gastric tissue, was associated with gastritis/erosion compared with healthy mucosa. It remains to be determined whether Cx43 has a role in the pathogenesis of disease due to H. pylori infection. However, most of the H. pylori isolated from Japanese gastric mucosa are VacA positive. Thus, VacA may participate in the generation of increased Cx43 in vivo. Interestingly, Liu et al. 72 reported that Cx43 expression decreases as gastric mucosa progresses to a precancerous lesion and then to cancer. These data are in agreement with our model, as the expected rate of epithelial apoptosis would generally decline during carcinogenesis. It has also been shown that during H. pylori-associated gastric cancer, the Cx43 gene promoter was hyper methylated, resulting in decreased Cx43 mRNA expression. 73 Our findings imply that VacA-increased Cx43 may contribute to tissue damage through epithelial apoptosis during H. pylori infection. Cx43 may thus be a potential therapeutic target. Reduction of Cx43 may have anti-inflammatory effects and inhibit the development of H. pylori VacA-induced tissue damage.

Purification of VacA
The toxin-producing H. pylori strain ATCC 49503 was the source of VacA for purification as previously described. 76 Assay for vacuolating activity Vacuolating activity was assessed using AZ-521 cells as previously described. 76 Briefly, cells (1 × 10 4 cells per well, 100 μl) were grown as monolayers in 96-well culture plates for 24 h in a 5% CO 2 atmosphere at 37°C. VacA was added, and cells were incubated at 37°C for the indicated times. To quantify vacuolating activity, uptake of neutral red into vacuoles was determined.

Subcellular fractionation
Subcellular fractionation was performed according to a previous study. 77 Briefly, AZ-521 cells (3 × 10 5 cells per six-well plate) were disrupted by Dounce homogenization in 100 μl of hypotonic buffer with 100 strokes, then centrifuged for 5 min at 800 × g at 4°C. The supernatant (total cell lysate fraction) was centrifuged for 15 min at 17 400 × g at 4°C. The supernatant (cytoplasmic fraction) was collected. The pellet was suspended in 50 μl of cell lysis buffer (50 mM Tris-HCl, pH 7.7, 0.15 M NaCl, 10% glycerol, 1% Triton X-100 (Tx) and protease inhibitor cocktail (Roche)) and then incubated for 30 min on ice. After centrifugation for 15 min at 17 400 × g at 4°C, the supernatant (Tx-soluble fraction) was collected and the pellet was solubilized with 50 μl of 1 × SDS sample buffer (Tx-insoluble fraction).

Triton extraction
Triton X-100 extraction of AZ-521 cells cultured on coverslips was performed as described previously. 78

Immunoprecipitation
Co-immunoprecipitation of conformationally changed Bax or Bak was performed as described previously. 79 Briefly, the indicated siRNAtransfected AZ-521 cells were treated with heat-inactivated VacA (iV) or VacA (V) for 6 h. Cells were solubilized with cell lysis buffer (10 mM HEPES, 150 mM NaCl, 1.5 mM MgCl 2 , 1 mM EGTA, 2% CHAPS, pH 7.4), centrifuged at 17 400 × g for 15 min at 4°C, and incubated with conformation-specific anti-Bax antibody (clone 3) (BD Biosciences) or anti-Bak antibody (Ab-2) (Calbiochem, San Diego, CA, USA) at 4°C for 3 h. Immunocomplexes were collected by incubation with protein G-Sepharose (Invitrogen), washed with cell lysis buffer three times, and dissolved in SDS-sample buffer. These samples were analyzed by SDS-PAGE in 15% gels, and transferred to PVDF membranes, which were then analyzed by immunoblotting using anti-Bax or anti-Bak antibodies (Cell Signaling Technology). After incubation with heat-inactivated VacA (iV) or VacA (V) for 8-10 h, cells were solubilized with RIPA buffer, centrifuged at 17 400 × g for 15 min at 4°C, and then incubated with anti-Cx43 monoclonal antibody (BD) overnight at 4°C. Immunocomplexes were washed with RIPA buffer three times, and dissolved in SDS-sample buffer. These samples were analyzed by SDS-PAGE in 15% gels, and transferred to PVDF membranes, which were then analyzed by immunoblotting using the indicated antibodies.

Real-time quantitative PCR analysis
Total RNA from AZ-521 cells (2 × 10 5 cells) was extracted by ISOGEN II (Wako) as described in the instruction manual. First-strand cDNA synthesis was performed with a PrimeScript II 1st strand cDNA Synthesis Kit (Takara Bio). Real-time quantitative PCR (qPCR) analysis was conducted with the fluorescent dye SYBR Green methodology using Power SYBR Green PCR Master Mix (ABI) and ABI Prism 7300 (Applied Biosystems Inc, Carlsbad, CA, USA). Primer pairs for the Cx43 and GAPDH genes were obtained from Takara. The PCR protocol was performed as described in the instruction manual: the reaction mix containing AmpliTaq Gold DNA polymerase was incubated for 10 min at 95°C, followed by denaturation for 15 s at 95°C with annealing and extension for 1 min at 60°C, for 45 cycles. The dissociation curve for each sample was analyzed to verify the specificity of each reaction. The relative mRNA expression levels of Cx43 genes were determined by the delta-delta Ct method and normalized to GAPDH expression.