Conversion of Helicobacter pylori CagA from senescence inducer to oncogenic driver through polarity-dependent regulation of p21

To investigate the effect of H. pylori CagA on epithelial cell proliferation, we inducibly expressed CagA in MKN28 human gastric epithelial cells using a tet-off system. As previously reported, CagA activated Erk MAP kinase but paradoxically inhibited cell proliferation, which was concomitantly associated with the accumulation of the CDK inhibitor p21 in cells (Fig. 1, A and B; Tsutsumi et al., 2003; Higashi et al., 2004; Murata-Kamiya et al., 2007). The growth-inhibitory activity of CagA was reproduced in AGS human gastric epithelial cells (Fig. S1, A and B). Knockdown of p21 by specific short hairpin (sh) RNA or small interfering (si) RNA abolished the ability of CagA to inhibit cell proliferation, indicating that elevated p21 was responsible for the CagA-mediated proliferation arrest (Fig. 1 C and Fig. S1 C). Treatment of cells with a MEK inhibitor U0126 also abrogated p21 accumulation by CagA (Fig. 1 D), whereas inhibition of PKC, PI-3 kinase, or PLC-γ, each of which can independently induce p21, did not have any effect on the CagA-mediated p21 accumulation (not depicted). Thus, CagA causes accumulation of p21 through Erk signaling. After exposure to CagA for 5 d, proliferation-arrested cells became flat and expressed senescence-associated β-galactosidase (Fig. 1 E). These results indicated that CagA expressed in nonpolarized epithelial cells aberrantly activates Erk signaling, which induces the accumulation of p21 and thereby causes senescence-like proliferation arrest.

The observations described in the previous section, in turn, indicated that CagA must have a mechanism that converts the response of host epithelial cells from growth inhibition to growth stimulation to exert its oncogenic action. During H. pylori infection in the stomach, CagA is delivered into the gastric mucosal monolayer comprised of epithelial cells with highly developed apical-basal polarity. We therefore sought to determine the effect of CagA on polarized epithelial cells, pathophysiologically relevant target cells for H. pylori. To this end, we used the Madin-Darby canine kidney (MDCK) II model epithelial monolayer. To prepare a polarized epithelial monolayer, we cultured MDCK cells on Transwell membrane filters for 3 d. In all experiments using polarized MDCK cells described in this paper, the establishment of apical-basal epithelial polarity was confirmed by staining cells with the tight junction marker ZO-1 (Fig. S2, representative staining images of polarized MDCK monolayers).

As previously described, expression of CagA in polarized MDCK cells led to the dissolution of tight junctions and loss of apical-basal polarity, followed by extrusion of CagA-expressing cells from the polarized monolayer (Amieva et al., 2003; Saadat et al., 2007). To our surprise, CagA-expressing cells that had extruded from the monolayer underwent multiple rounds of cell divisions without showing p21 accumulation (Fig. 2, A and B). Both WT CagA and phosphorylation-resistant (PR) CagA were capable of inducing DNA synthesis when expressed in polarized epithelial cells, although the degree of DNA synthesis induced by PR CagA was weaker than that induced by WT CagA (Fig. 2 C). To determine whether the observed mitogenesis by CagA is cell polarity dependent or cell type dependent, we transiently expressed CagA in nonpolarized MDCK cells, which had been prepared by plating MDCK cells in a standard cell culture condition without using Transwell and found that, as in MKN28 and AGS cells, CagA induced p21 and inhibited cell proliferation independently of cell density (Fig. S3). The possibility that the Transwell culture on its own modified cellular response to CagA was excluded from the observation that expression of CagA in Transwell-cultured MKN28 cells, which cannot develop tight junctions and apical-basal epithelial polarity, induced p21 (Fig. S4). In contrast to WT CagA, CagA-ΔCM/MKI, which does not bind PAR1 and thus cannot disrupt epithelial polarity, failed to provoke cell extrusion and subsequent mitogenesis when expressed in polarized MDCK cells (Fig. 2 C). Accordingly, the mitogenic effect of CagA is epithelial-polarity context dependent. Treatment of polarized MDCK cells with U0126 abolished extrusion of CagA-expressing cells from the polarized monolayer and subsequent mitogenesis, indicating that, like PAR1 inhibition, CagA-induced Erk activation is required for both extrusion and proliferation (Fig. 2 D). We also investigated the polarity-dependent effects of CagA on cell proliferation using MDCK cells that inducibly express WT CagA (MDCK-WT-CagA cells) or PR CagA (MDCK-PR-CagA cells) by a tet-on system (Fig. S5). As expected, induction of CagA in nonpolarized MDCK cells caused p21-mediated cell senescence, whereas CagA, either WT or phosphorylation-resistant mutant, inducibly expressed in polarized epithelial cells initiated DNA synthesis (Fig. S5). Collectively, these observations indicated that aberrant Erk signaling, which is activated via both tyrosine phosphorylation-dependent and -independent CagA activities, can induce opposite cellular responses, senescence, and forced mitogenesis, depending on the polarity status of the epithelial cell to which CagA is delivered.

It has been reported that CagA and the vacuolating toxin VacA, another H. pylori virulence factor, down-regulate each other's effects on epithelial cells (Yokoyama et al., 2005; Argent et al., 2008; Tegtmeyer et al., 2009). In particular, VacA has been shown to inhibit Erk signaling by inactivating epidermal growth factor receptor (EGF-R) and HER2/neu kinases, although the underlying mechanisms remain unknown (Tegtmeyer et al., 2009). To investigate whether VacA influences the polarity context-dependent action of CagA on cell proliferation, we expressed CagA in polarized or nonpolarized MDCK cells in the presence or absence of 5 µg/ml of acid-activated VacA. The VacA treatment potently induced vacuolation of MDCK cells but did not influence the polarity-dependent regulation of p21 by CagA (Fig. 2, E–G). These observations argue against the idea that VacA directly modulates the effect of CagA on proliferation of both polarized and nonpolarized MDCK epithelial cells. Because infection of polarized MDCK cells with H. pylori was extremely difficult (unpublished data), we could not investigate the polarity-dependent effect of VacA on CagA by infection experiments with MDCK cells.

Extrusion of CagA-expressing cells from the polarized epithelial monolayer was concomitantly associated with the morphological transition of epithelial cells from a polarized state to an invasive phenotype, a cellular change characteristic of EMT (Fig. 2 A; Bagnoli et al., 2005). To investigate CagA-induced EMT in more detail, we examined the expression of mesenchymal markers after expression of CagA in MDCK cells and found that CagA significantly increased levels of mesenchymal proteins such as vimentin and fibronectin (Fig. 3, left). However, CagA expression did not down-regulate epithelial markers such as E-cadherin, α-catenin, β-catenin, and γ-catenin (Fig. 3, middle). Furthermore, there was no up-regulation of EMT-inducing transcription factors, such as Twist and Snail, in cells expressing CagA (Fig. 3, right). Hence, CagA-expressing cells that have undergone morphological EMT simultaneously express both epithelial and mesenchymal markers.

We next investigated the mechanism by which CagA-dependent p21 accumulation is abolished in an epithelial-polarity context-dependent manner. It has been reported that RhoA counteracts induction of p21 by oncogenic Ras signaling (Adnane et al., 1998; Olson et al., 1998; Sahai et al., 2001; Liberto et al., 2002; Coleman et al., 2006). We therefore postulated that RhoA is involved in the inhibition of CagA-dependent p21 accumulation. To test this idea, we treated polarized MDCK cells with a membrane-permeable RhoA inhibitor, Clostridium botulinum C3 transferase. The C3 transferase treatment caused accumulation of p21 in polarized epithelial cells expressing CagA (Fig. 4, A and B). Furthermore, RhoA was activated upon expression of CagA in polarized epithelial cells but not in nonpolarized epithelial cells, as determined by a GST pull-down assay (Fig. 4, C and D) or the use of a RhoA-specific fluorescence resonance energy transfer (FRET) probe (Fig. 4 E). From these observations, we concluded that activation of RhoA overcomes CagA/Erk-dependent accumulation of p21.

The observations described in the previous section prompted us to investigate the mechanism underlying RhoA activation by CagA. RhoA activity is controlled via the interplay between activators, GEFs, and inhibitors, GAPs (GTPase-activation proteins). Among various RhoA-specific GEFs, GEF-H1 was of particular interest because the GEF activity is regulated by polarity-dependent mechanisms (Birkenfeld et al., 2008). In nonpolarized MDCK cells, GEF-H1 was distributed to the cytoplasm, where it was associated with microtubules (Fig. 5 A). This GEF-H1–microtubules interaction has been reported to inhibit RhoA-specific GEF activity of GEF-H1 (Krendel et al., 2002; Birkenfeld et al., 2008; Chang et al., 2008). In polarized MDCK cells, however, GEF-H1 was translocalized to the membrane and was concentrated to the tight junctions. At the tight junctions, GEF-H1 specifically interacted with a GEF-H1 inhibitor, cingulin, as previously reported (Fig. 5 B; Aijaz et al., 2005; Birkenfeld et al., 2008). Accordingly, we investigated the effect of CagA on the level of GEF-H1–cingulin complex in polarized epithelial cells. To this end, we immunoprecipitated GEF-H1 in lysates prepared from polarized MDCK cells with or without CagA expression and then immunoblotted the GEF-H1 immunoprecipitates with an anti-cingulin antibody. As shown in Fig. 5 C (top), the amount of GEF-H1-bound cingulin was markedly decreased when CagA was coexpressed in polarized MDCK cells. In contrast, there was no change in the amount of GEF-H1–bound cingulin by CagA in nonpolarized MDCK cells (Fig. 5 C, bottom). The observations indicated that, upon expression in polarized epithelial cells, CagA disrupts the GEF-H1–cingulin complex and, thereby, liberates GEF-H1 from the constraints by cingulin. To investigate whether junctional and polarity defects caused by CagA are responsible for disruption of the GEF-H1–cingulin complex, we destroyed the apical junctional complex of polarized MDCK cells by calcium depletion. Like CagA, calcium depletion led to dissociation of cingulin from GEF-H1, which was concomitantly associated with RhoA activation as determined by a pulldown assay (Fig. 5 D). These results indicated that junctional and polarity defects, whether they are induced by CagA or calcium depletion, cause disruption of the GEF-H1–cingulin complex and subsequent activation of GEF-H1 and then RhoA.

To consolidate the importance of GEF-H1 in the suppression of p21 expression, we inhibited GEF-H1 expression by specific siRNA in polarized MDCK cells and found that the GEF-H1 knockdown led to the accumulation of p21 in cells expressing CagA (Fig. 6 A and Fig. S6). Conversely, coexpression of GEF-H1 with CagA in nonpolarized MDCK cells abolished CagA-mediated p21 accumulation (Fig. 6 B). From these observations, we concluded that specific activation of GEF-H1 by CagA expressed in polarized epithelial cells stimulates RhoA, which in turn suppresses p21 accumulation.

We next wished to gain an insight into the mechanism by which CagA expressed in polarized epithelial cells enforces mitogenesis even after cells become nonpolarized. As described in the previous section, expression of CagA under a polarized condition activates GEF-H1, whereas CagA expression in a nonpolarized condition fails to do so. Because GEF-H1–mediated RhoA activation is required for the suppression of p21 accumulation, there must be a mechanism that keeps GEF-H1 in its active form when CagA was introduced into polarized epithelial cells. In this regard, GEF-H1 has been reported to bind PAR1d, a member of the PAR1/MARK family (Brajenovic et al., 2004). We confirmed and extended the observation by showing that all of the PAR1 isoforms (PAR1a-d) interact with GEF-H1 independent of cingulin binding (Fig. 7 A). Because CagA also binds to these PAR1 isoforms (Saadat et al., 2007; Lu et al., 2009), we hypothesized that CagA can associate with GEF-H1 using PAR1 as an adaptor. Upon triple transfection in COS-7 cells, we found that CagA forms a physical complex with GEF-H1 via PAR1 (Fig. 7, B and C). We also found that RhoA G17A, which specifically binds to the active form of GEF-H1 (García-Mata et al., 2006), pulled down GEF-H1 together with CagA, indicating that GEF-H1 was active in the heterotrimeric complex (Fig. 7 D). Because CagA is tethered to the plasma membrane, the CagA–PAR1–GEF-H1 heterotrimeric complex should prevent translocation of GEF-H1 to the cytoplasm, where it becomes inactivated. Consistent with this notion, GEF-H1 was colocalized with CagA at the membrane in CagA-expressing cells that had been expelled from the polarized monolayer (Fig. 7 E). These results indicate that sustained membrane localization of active GEF-H1 by CagA enables continued RhoA activation, which inhibits p21 accumulation, during and after transition of epithelial cells from a polarized state to nonpolarized state.

RhoA has two well known downstream effectors: Rho-associated kinase (ROCK) and formin homology protein mDia. To further delineate the mechanism by which active RhoA inhibits CagA-dependent p21 accumulation, we examined the effect of Y-27632, a specific inhibitor of ROCK. Treatment of polarized MDCK cells with Y-27632 led to the CagA-dependent accumulation of p21, indicating that ROCK is the downstream effector of RhoA that mediates p21 suppression (Fig. 8 A). We next investigated the mechanism underlying ROCK-mediated suppression of p21 accumulation. Previous studies demonstrated that ROCK stimulates c-Myc, which has been reported to effectively inhibit p21 expression (Watnick et al., 2003; Liu et al., 2009b). Consistent with this, the level of phospho-c-Myc, the active form of c-Myc, was markedly increased upon expression of CagA in polarized MDCK cells (Fig. 8 B and Fig. S7 A). This increase in the active form of c-Myc was inhibited by Y-27632 (Fig. 8 C and Fig. S7 B). In contrast, expression of CagA in nonpolarized MDCK cells induced p21 accumulation but failed to activate c-Myc (Fig. 8 D). To consolidate the role of c-Myc in CagA-dependent inhibition of p21 accumulation, we specifically inhibited c-Myc expression by siRNA in polarized MDCK cells. We found that down-regulation of c-Myc causes accumulation of p21 in CagA-expressing MDCK cells (Fig. 8 E). These observations indicated that CagA expressed in polarized epithelial cells activates the GEF-H1–RhoA–ROCK cascade, which in turn stimulates c-Myc to repress CagA-dependent p21 accumulation.

It has been reported that RhoA-mediated inhibition of p21 accumulation involves posttranscriptional mechanisms (Coleman et al., 2006). Recently, microRNAs have emerged as key posttranscriptional regulators for gene expression (Bartel, 2004). Intriguingly, c-Myc has been shown to induce the miR-17-92 cluster, which contains miR-17 and miR-20a that target p21 mRNA (Petrocca et al., 2008; Inomata et al., 2009). We therefore hypothesized that c-Myc–regulated miR-17-92 is involved in the regulation of p21 expression in cells expressing CagA. Quantitative RT-PCR analysis revealed that expression of CagA in polarized MDCK cells (Fig. 9 A) or depolarization of MDCK cells by calcium depletion (Fig. S7 C) elicits significant increases in the amounts of miR-17 and miR-20a, which were abolished by treatment of cells with the ROCK inhibitor Y-27632. Again, the presence of VacA did not influence induction of these microRNAs by CagA (Fig. 9 B). To consolidate the role of c-Myc in CagA-dependent elevation of the microRNAs, we specifically inhibited c-Myc expression by siRNA in polarized MDCK cells. We found that down-regulation of c-Myc abolishes induction of miR-17 or miR-20a by CagA (Fig. 9 C). Conversely, inhibition of miR-17 and miR-20a expression by antisense oligonucleotides in polarized epithelial cells led to the accumulation of p21 by CagA (Fig. 9 D). These observations indicated that c-Myc–induced miR-17 and/or miR-20a plays a dominant role in the inhibition of CagA-dependent p21 accumulation.

In the present study, we found that H. pylori CagA expressed in nonpolarized epithelial cells aberrantly activates growth-promoting Erk signaling, which exerts an oncogenic stress that gives rise to p21 accumulation and subsequent cellular senescence. In striking contrast, expression of CagA in polarized epithelial cells not only elicits Erk activation but also generates a co-lateral signal that overrides the oncogenic stress-induced cell senescence (Fig. S8). Hence, a single bacterial protein creates an integrated biological output that mimics oncogene collaboration (i.e., ras and myc collaboration) to bypass senescence and yet induce forced mitogenesis in an epithelial-polarity context-dependent manner.

Deregulated Erk activation by gain-of-function mutations in oncogenes, especially those constituting the Ras–Erk pathway, leads to premature senescence in primary cells through induction of CDK inhibitors such as p21 and p16 (Sewing et al., 1997; Woods et al., 1997; Roovers and Assoian, 2000). Oncogene-induced senescence has been considered to act as a cell-autonomous protective response against neoplastic transformation (Ben-Porath and Weinberg, 2005; Zhang, 2007). Accumulation of p21 is primarily a result of Erk-mediated phosphorylation and activation of Sp1 (Kim and Lim, 2009; Tan and Khachigian, 2009), which directs p21 transcription. Like endogenous oncogenes, H. pylori CagA aberrantly activates Erk signaling by deregulating cellular proteins, such as SHP-2, and thereby exerts oncogenic stress when expressed in nonpolarized epithelial cells, hampering its potential as a mitogenic oncoprotein (Tsutsumi et al., 2003; Higashi et al., 2004; Murata-Kamiya et al., 2007).

In the H. pylori–infected stomach, CagA is delivered into the polarized epithelial monolayer via type IV secretion. Upon delivery, CagA deregulates Erk signaling and, at the same time, inhibits the polarity-regulating kinase PAR1/MARK, resulting in disruption of tight junctions and loss of apical-basal epithelial polarity. As a consequence, CagA-expressing cells extrude from the polarized epithelial monolayer (Amieva et al., 2003; Bagnoli et al., 2005; Saadat et al., 2007). Similar cell extrusion occurs when Ras- or Src-transformed cells are surrounded by polarized normal epithelial cells both in vitro and in vivo (Hogan et al., 2009; Kajita et al., 2010), indicating that CagA-expressing cells have acquired malignant properties shared by Ras- or Src-transformed cells. Intriguingly, CagA-positive cells that have expelled from the polarized monolayer start DNA synthesis and subsequent cell multiplication in response to CagA-activated Erk signaling without causing p21 accumulation, even after they become nonpolarized (Fig. 2). This indicates that, during transition from a polarized state to a nonpolarized state, CagA-expressing epithelial cells acquire the ability to override Erk-dependent p21 accumulation independent of cell polarity status. Once established, this p21-inhibitory mechanism is kept active until the CagA level decreases below a certain threshold. Although the mechanism underlying this phenomenon needs further investigation, the results of the present study suggested that sustained membrane localization of active GEF-H1 plays an important role in this process. In polarized epithelial cells, GEF-H1 is recruited to the membrane, whereas in nonpolarized cells it is distributed to the cytoplasm and its activity is inhibited by binding with microtubules. H. pylori–delivered CagA localizes to the inner surface of the plasma membrane by interacting with phosphatidylserine (Murata-Kamiya et al., 2010). In polarized epithelial cells, membrane-localized CagA binds and inhibits PAR1 and thereby provokes junctional and polarity defects, which in turn elicit the activation of GEF-H1 via its dissociation from the inhibitor cingulin. Released GEF-H1 then interacts with the membrane-tethered CagA through PAR1, keeping active GEF-H1 localized to the membrane. Sustained membrane localization of active GEF-H1, which is achieved only when CagA has been delivered into polarized epithelial cells, may enable continued inhibition of p21 accumulation even after cells become nonpolarized.

Activated Erk signaling also plays an essential role in CagA-induced cell elongation, known as the hummingbird phenotype, in nonpolarized gastric epithelial cells such as AGS cells (Higashi et al., 2002, 2004). Notably, H. pylori infection can activate Erk signaling independent of CagA, probably via stimulation of growth factor receptors such as EGF-R and HER2/neu (Keates et al., 2001; Wallasch et al., 2002; Du et al., 2007; Tegtmeyer et al., 2009). This H. pylori infection–activated Erk signaling acts together with CagA-activated Erk signaling to enhance induction of the hummingbird phenotype by CagA (Tegtmeyer et al., 2009). These observations indicate that infection with cagA-positive H. pylori causes a higher level of Erk activation than that achieved by CagA alone, resulting in greater accumulation of p21 in nonpolarized cells and stronger mitogenesis in polarized epithelial cells. It has also been reported that CagA and VacA down-regulate each other's effects on epithelial cells. VacA diminishes induction of the hummingbird phenotype upon cagA-positive H. pylori infection by specific inhibition of EGF-R and neu-dependent Erk activation (Argent et al., 2008; Tegtmeyer et al., 2009). Notably, however, VacA does not directly inhibit the ability of CagA to induce the hummingbird phenotype (Yokoyama et al., 2005). Consistent with these notions, we found in this paper that VacA does not influence the effect of CagA on p21 expression in both polarized and nonpolarized MDCK cells.

CagA-expressing cells expelled from the polarized epithelial monolayer show morphological transformation to an invasive mesenchymal phenotype, a phenomenon known as EMT. Initiation of the EMT program is usually associated with loss of functional E-cadherin, which in many cases is achieved through transrepression by EMT-inducing transcription factors such as Twist, Snail, and Zeb1 (Acloque et al., 2009; Kalluri and Weinberg, 2009). However, recent studies also suggest that disorganization of tight junctions could trigger the EMT program as well (Barrios-Rodiles et al., 2005; Ozdamar et al., 2005). Accordingly, CagA-mediated disruption of tight junctions may trigger the EMT program in polarized epithelial cells. Reported interaction of CagA with E-cadherin, which impairs E-cadherin activity, may also contribute to the progression of EMT (Murata-Kamiya et al., 2007; Kurashima et al., 2008; Oliveira et al., 2009). CagA-activated Erk signaling could further promote EMT through disassembly of adherens junctions as well as elevated cell motility (Zavadil et al., 2001). During EMT, CagA-expressing cells become positive for mesenchymal markers, although they neither down-regulate epithelial markers nor induce EMT-inducing transcription factors. Hence, CagA-triggered EMT represents an aberrantly dedifferentiated/transdifferentiated cellular status.

Recent studies have indicated a relationship among p21, senescence, and EMT. EMT-inducing transcription factor Twist inhibits p21 (Ansieau et al., 2008). Conversely, p21 reverses Twist-mediated E-cadherin repression (Li et al., 2009). Mice deficient in another EMT-inducing transcription factor Zeb1 exhibit an EMT-like change and Zeb1-null mouse embryonic fibroblasts show accumulation of p21 (Liu et al., 2008). In addition, EMT confers resistance to oncogene-induced cell senescence by suppressing accumulation of p21 and/or p16 (Ansieau et al., 2008), whereas elevated p21 attenuates Ras- and c-Myc–dependent induction of EMT (Liu et al., 2009a). These observations strongly suggest that CDK inhibitors, such as p21, act as a molecular switch that determines cell fate, either senescence or EMT, in response to oncogenic insults. A key question is, therefore, how H. pylori CagA regulates p21 in an epithelial polarity-dependent manner. We found in this work that RhoA is a key determinant in this process. We also found that the GEF-H1–cingulin complex plays a critical role in the activation of RhoA after expression of CagA in polarized epithelial cells. GEF-H1 is a RhoA-specific GDP/GTP exchanging factor whose activity is inhibited by interaction with cingulin (Aijaz et al., 2005; Birkenfeld et al., 2008). During apical-basal polarization of epithelial cells, the GEF-H1–cingulin complex is formed and concentrated to the tight junctions. Expression of CagA in polarized epithelial cells elicits junctional and polarity defects, which enforce dissociation of cingulin from the GEF-H1–cingulin complex. Being liberated from constraints by cingulin, GEF-H1 stimulates RhoA, which results in the suppression of p21 accumulation (Auer et al., 1998; Olson et al., 1998).

Our findings also provide an insight into the mechanisms underlying suppression of p21 accumulation by activated RhoA. GEF-H1–activated RhoA stimulates the Rho kinase ROCK, which in turn increases the active form of c-Myc. It has been reported that c-Myc inhibits p21 transcription via multiple distinct mechanisms (Wanzel et al., 2003). Notably, however, inhibition of p21 by RhoA involves posttranscriptional mechanisms (Coleman et al., 2006). In fact, we found that c-Myc induces microRNAs—miR-17 and miR-20a—that target p21 mRNA and that these microRNAs plays a dominant role in the inhibition of CagA/Erk-dependent p21 accumulation when CagA is expressed in polarized epithelial cells.

A series of studies on Drosophila melanogaster tumor suppressor genes demonstrated a role for epithelial polarity in preventing tumor development in flies (Tapon, 2003; Bilder, 2004). Consistently, loss of epithelial polarity is a hallmark of carcinomas in mammals. Nevertheless, the actual contribution of polarity defects in mammalian carcinogenesis has remained uncertain. The present work has demonstrated a causal relationship between epithelial polarity and proliferation control, providing evidence for a tumor-suppressive role of polarized epithelial cells. Disruption of cell–cell junctions and subsequent loss of epithelial polarity liberate cells from growth-inhibitory cues through activation of the GEF-H1–RhoA–ROCK–c-Myc–microRNA–p21 signaling axis, a long-sought signaling pathway which directly links epithelial polarity and the cell cycle. We suspect that deregulation of the identified GEF-H1–RhoA–ROCK–c-Myc–microRNA–p21 axis is broadly and substantially involved in the neoplastic transformation of epithelial cells, other than gastric epithelial cells.

CagA-mediated increase in the number of abnormal epithelial cells that have undergone EMT would have a significant impact on the physiological functions of the stomach mucosa such as gastric acid secretion. For H. pylori, reduced acidity should be of great advantage for long-term colonization in the stomach. Meanwhile, Mani et al. (2008) recently reported that EMT allocates epithelial cells with malignant characteristics, indicating that CagA-expressing epithelial cells that have undergone EMT acquire malignant traits. Accordingly, an unexpected adverse effect of CagA, which primarily acts for the better life of H. pylori, is that it allows continuous generation and expansion of a cancer-predisposed cell population in the H. pylori–infected stomach.

MKN28, AGS, COS-7, and MDCK (MDCK II) cells were cultured as previously described (Higashi et al., 2002; Murata-Kamiya et al., 2007; Saadat et al., 2007). WT-A10 cell is an MKN28-derived stable transfectant clone that inducibly expresses CagA using a tet-off system (Murata-Kamiya et al., 2007). The p21sh-6-12 cell is a WT-A10–derived clone made by stable transfection of pSUPER-p21, an expression vector for p21-specific shRNA. MDCK cells were grown on Transwell filters (Corning) to make polarized epithelial monolayer. MDCK cells that inducibly express CagA were established using Tet-on Advanced Inducible Gene Expression System (Takara Bio Inc.). In brief, MDCK cells were transfected with pTet-On-Advanced Vector and were selected in the presence of 800 µg/ml G418. Transfectants carrying pTet-On-Advanced were then transfected with pTRE-tight-CagA and a drug-resistant gene (pBabe-puro) and were selected in the presence of 1 µg/ml puromycin. MDCK transfectants that inducibly express in the presence of 0.2 µg/ml of a tetracycline analogue, Dox, were single cell cloned by limiting dilution method.

Mammalian expression vectors for WT CagA, PR-CagA, and a CagA mutant that lacks the PAR1/MARK-interacting sequence (CagA-ΔCM/MKI) were described previously (Higashi et al., 2002; Saadat et al., 2007). Human cDNA for GEF-H1 (Fujishiro et al., 2008) was provided by M. Kohno (Nagasaki University, Nagasaki, Japan). An expression vector for a RhoA-specific FRET probe, pRaichu-RhoA (1298x; Yoshizaki et al., 2003), was provided by M. Matsuda (Kyoto University, Kyoto, Japan). Plasmids were transfected into cells as previously described (Higashi et al., 2002). Recombinant adenoviruses that express a RhoA FRET probe, WT CagA, and β-galactosidase were generated using ViraPower Adenoviral Expression System (Invitrogen).

Anti-HA (3F10; Roche), anti-Flag (Sigma-Aldrich), anti-c-Myc (9E10; Santa Cruz Biotechnology, Inc.), anti-T7 (sc-7270; Santa Cruz Biotechnology, Inc.), anti-RhoA (sc-179; Santa Cruz Biotechnology, Inc.), anti-RhoA (Millipore), anti-actin (sc-1615; Santa Cruz Biotechnology, Inc.), anti-ERK1 (sc-93; Santa Cruz Biotechnology, Inc.), anti–p-ERK (sc-7383; Santa Cruz Biotechnology, Inc.), anti-p21 (sc-397; Santa Cruz Biotechnology, Inc.), anti-bromodeoxyuridine (Roche), anti-CagA (AUSTRAL Biologicals), anti-HA (6E2; Cell Signaling Technology), anti–phospho-c-Myc (Cell Signaling Technology), anti–c-Myc (Cell Signaling Technology), anti–GEF-H1 (B4/7; Hycult Biotech), anti–α-tubulin (sc-53029; Santa Cruz Biotechnology, Inc.), anti–Snail 1 (sc-28199; Santa Cruz Biotechnology, Inc.), anti-Twist (3E11; Abnova), anti–ZO-1 (Invitrogen), anti–α-catenin (BD), anti–β-catenin (BD), anti–γ-catenin (BD), anti-vimentin (V9; Thermo Fisher Scientific), anti-fibronectin (BD), anti–E-cadherin (BD), anti–N-cadherin (BD), and anti-cingulin (Invitrogen) antibodies were used in this study. A rabbit anti-PAR1b antibody was provided by S. Ohno (Yokohama City University, Yokohama City, Japan). MEK inhibitor U0126 and its control U0124 (EMD) were used at the final concentration of 25 µM. Cell permeable C3 transferase from Clostridium botulinum (Cytoskeleton, Inc.) was used at the final concentration of 5 µg/ml. ROCK inhibitor Y-27632 was purchased from EMD and was used at the final concentration of 50 µM. Immunoprecipitation and immunoblotting were performed as previously described (Higashi et al., 2002).

Cells were collected by trypsinization, washed with PBS, and fixed in 70% cold ethanol at −20°C overnight. After centrifugation, cells were washed with PBS and treated for 20 min at 37°C with 0.5 mg/ml RNase A. Cells were stained at room temperature with 50 µg/ml propidium iodide and subjected to flow cytometric analysis using FACScan and Cell Quest software (BD).

Cells were fixed and permeabilized as previously described (Higashi et al., 2002). The cells were then treated with primary antibodies and visualized with Alexa Fluor conjugate secondary antibodies (Invitrogen). Images were captured with a confocal microscope system (Olympus).

Nonpolarized or polarized MDCK cells were transfected with expression vectors. At 12 h after transfection, 10 µM BrdU was added to the culture and cells were labeled with BrdU for an additional 15 h. After fixation and permeabilization, cells were treated with 5 U/µl DNase for 1 h at 37°C and were stained with an anti-BrdU antibody.

MDCK II cells were transfected with pRaichu-RhoA together with expression vector. At 24 h after transfection, cells were washed and lysed as previously described (Yoshizaki et al., 2003). Cell lysates were centrifuged and transferred into 96-well plates. Fluorescence was measured with a fluorescence spectrophotometer (LS 50; PerkinElmer) using an excitation wavelength of 430 nm. The ratios of emission at 530/475 nm were calculated. In several experiments, polarized MDCK II cells were cultured in calcium-free medium for 24 h and then coinfected with recombinant adenoviruses that transduce a FRET probe for RhoA and CagA-WT or β-galactosidase. At 24 h after infection, cells were lysed and fluorescence was measured with a fluorescence spectrophotometer.

RhoA activity was determined using Rho Assay Reagent kit according to the manufacturer's instruction (Millipore). Quantitation of RhoA bands obtained by anti-RhoA immunoblotting was performed using a luminescence image analyzer (LAS-4000 IR multi color; Fujifilm).

GEF-H1–specific siRNA-1 (5′-GCUUUACAAGGAGCUGUAUTT-3′) and GEF-H1–specific siRNA-2 (5′-CCAGGAGAAAGGGAUGUTT-3′) or c-Myc–specific siRNA-1 (5′-GAGGCGAACACACAACGUCTT-3′) and c-Myc–specific siRNA-2 (5′-GCGGCGAGAACAGUUGAAATT-3′) were synthesized. p21-specific siRNA (Hs-CDKN1A-5 FlexiTube siRNA) was purchased from QIAGEN. Each 100-pmol siRNA was transfected into cells. At 12 h after transfection, cells were reseeded on Transwell. After 3 d of culture, cells were transfected with plasmid vectors. At 27 h after transfection, cells were fixed and stained with anti-p21 antibody.

CagA-inducible MKN28 or CagA-inducible MDCK cells were cultured for 5 d in the presence or absence of Dox. Cells were then washed with PBS, fixed with 3% formaldehyde, and incubated overnight with staining buffer (1 mg/ml X-Gal, 40 mM citric acid/sodium phosphate, pH 6.0, 150 mM NaCl, 2 mM MgCl₂, 5 mM K₃Fe(CN)₆, and 5 mM K₄Fe(CN)₆) for detecting SA-β-galactosidase activity.

Expression of mature microRNAs was measured using TaqMan MicroRNA Assays (Applied Biosystems). In brief, total RNA was extracted with the use of TRIzol reagent (Invitrogen) according to the manufacturer's protocol. Reverse transcription was conducted with TaqMan MicroRNA Reverse Transcription kit (Applied Biosystems) using specific stem-loop primer for miR-17, miR-20a, and RNU6B (control). Real-time PCR was performed with LightCycler 480 (Roche) using LightCycler Taqman master (Roche) and TaqMan MicroRNA assay system (Applied Biosystems). Expression of microRNAs was normalized using U6 snRNA (RNU6B) as a control.

Antisense oligonucleotide against miR-17 (5′-CTACAAGTGCCTTCACTGCAGT-3′), antisense oligonucleotide against miR-20a (5′-CTACCTGCACTATAAGCACTTTA-3′), and control oligonucleotide (5′-TAACGTCACTTCGACTGAAGTGCT-3′) were synthesized with modification of LNA (locked nucleic acid; 2′-O and 4′-C methylene bridge modification) at G and C residues (GeneDesign Inc.). Polarized MDCK cells were cotransfected with an antisense oligonucleotide and an expression vector. 27 h after transfection, cells were fixed and stained with anti-p21 and anti-HA antibodies.

Statistical analyses were performed by the Student's t test.

Fig. S1 shows inhibition of cell proliferation by CagA in nonpolarized AGS gastric epithelial cells. Fig. S2 shows representative staining images of nonpolarized or polarized MDCK cells. Fig. S3 shows inhibition of proliferation by CagA in nonpolarized MDCK cells. Fig. S4 shows inhibition of proliferation by CagA in MKN28 cells cultured on Transwell filters. Fig. S5 shows induction of proliferation by CagA in polarized MDCK cells. Fig. S6 shows requirement of GEF-H1 for the inhibition of CagA-mediated p21 accumulation in polarized epithelial cells. Fig. S7 shows induction of p21-targeting microRNAs upon calcium depletion by activation of the ROCK–c-Myc pathway. Fig. S8 summarizes a model for epithelial polarity-dependent cellular response to CagA.

We thank Drs. Michiyuki Matsuda, Jun-ichi Miyazaki, Michiaki Kohno, and Shigeo Ohno for plasmids and antibodies. We also thank Dr. Tadatsugu Taniguchi for experimental help.

This work was supported by Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan and by the Mitsubishi Foundation Research Grant.

The authors declare that they have no conflicting financial interests.