Role of Partitioning-defective 1/Microtubule Affinity-regulating Kinases in the Morphogenetic Activity of Helicobacter pylori CagA*

Helicobacter pylori CagA plays a key role in gastric carcinogenesis. Upon delivery into gastric epithelial cells, CagA binds and deregulates SHP-2 phosphatase, a bona fide oncoprotein, thereby causing sustained ERK activation and impaired focal adhesions. CagA also binds and inhibits PAR1b/MARK2, one of the four members of the PAR1 family of kinases, to elicit epithelial polarity defect. In nonpolarized gastric epithelial cells, CagA induces the hummingbird phenotype, an extremely elongated cell shape characterized by a rear retraction defect. This morphological change is dependent on CagA-deregulated SHP-2 and is thus thought to reflect the oncogenic potential of CagA. In this study, we investigated the role of the PAR1 family of kinases in the hummingbird phenotype. We found that CagA binds not only PAR1b but also other PAR1 isoforms, with order of strength as follows: PAR1b > PAR1d ≥ PAR1a > PAR1c. Binding of CagA with PAR1 isoforms inhibits the kinase activity. This abolishes the ability of PAR1 to destabilize microtubules and thereby promotes disassembly of focal adhesions, which contributes to the hummingbird phenotype. Consistently, PAR1 knockdown potentiates induction of the hummingbird phenotype by CagA. The morphogenetic activity of CagA was also found to be augmented through inhibition of non-muscle myosin II. Because myosin II is functionally associated with PAR1, perturbation of PAR1-regulated myosin II by CagA may underlie the defect of rear retraction in the hummingbird phenotype. Our findings reveal that CagA systemically inhibits PAR1 family kinases and indicate that malfunctioning of microtubules and myosin II by CagA-mediated PAR1 inhibition cooperates with deregulated SHP-2 in the morphogenetic activity of CagA.

Infection with Helicobacter pylori strains bearing cagA (cytotoxin-associated gene A)-positive strains is the strongest risk factor for the development of gastric carcinoma, the second leading cause of cancer-related death worldwide (1)(2)(3). The cagA gene is located within a 40-kb DNA fragment, termed the cag pathogenicity island, which is specifically present in the genome of cagA-positive H. pylori strains (4 -6). In addition to cagA, there are ϳ30 genes in the cag pathogenicity island, many of which encode a bacterial type IV secretion system that delivers the cagA-encoded CagA protein into gastric epithelial cells (7)(8)(9)(10). Upon delivery into gastric epithelial cells, CagA is localized to the plasma membrane, where it undergoes tyrosine phosphorylation at the C-terminal Glu-Pro-Ile-Tyr-Ala motifs by Src family kinases or c-Abl kinase (11)(12)(13)(14). The C-terminal Glu-Pro-Ile-Tyr-Ala-containing region of CagA is noted for the structural diversity among distinct H. pylori isolates. Oncogenic potential of CagA has recently been confirmed by a study showing that systemic expression of CagA in mice induces gastrointestinal and hematological malignancies (15).
When expressed in gastric epithelial cells, CagA induces morphological transformation termed the hummingbird phenotype, which is characterized by the development of one or two long and thin protrusions resembling the beak of the hummingbird. It has been thought that the hummingbird phenotype is related to the oncogenic action of CagA (7, 16 -19). Pathophysiological relevance for the hummingbird phenotype in gastric carcinogenesis has recently been provided by the observation that infection with H. pylori carrying CagA with greater ability to induce the hummingbird phenotype is more closely associated with gastric carcinoma (20 -23). Elevated motility of hummingbird cells (cells showing the hummingbird phenotype) may also contribute to invasion and metastasis of gastric carcinoma.
In host cells, CagA interacts with the SHP-2 phosphatase, C-terminal Src kinase, and Crk adaptor in a tyrosine phosphorylation-dependent manner (16,24,25) and also associates with Grb2 adaptor and c-Met in a phosphorylation-independent manner (26,27). Among these CagA targets, much attention has been focused on SHP-2 because the phosphatase has been recognized as a bona fide oncoprotein, gain-of-function mutations of which are found in various human malignancies (17,18,28). Stable interaction of CagA with SHP-2 requires CagA dimerization, which is mediated by a 16-amino acid CagA-multimerization (CM) 2 sequence present in the C-terminal region of CagA (29). Upon complex formation, CagA aberrantly activates SHP-2 and thereby elicits sustained ERK MAP kinase activation that promotes mitogenesis (30). Also, CagA-activated SHP-2 dephosphorylates and inhibits focal adhesion kinase (FAK), causing impaired focal adhesions. It has been shown previously that both aberrant ERK activation and FAK inhibition by CagA-deregulated SHP-2 are involved in induction of the hummingbird phenotype (31).
In polarized epithelial cells, CagA disrupts the tight junctions and causes loss of apical-basolateral polarity (39,40). This CagA activity involves the interaction of CagA with PAR1b/MARK2 (19,41). CagA directly binds to the kinase domain of PAR1b in a tyrosine phosphorylation-independent manner and inhibits the kinase activity. Notably, CagA binds to PAR1b via the CM sequence (19). Because PAR1b is present as a dimer in cells (42), CagA may passively homodimerize upon complex formation with the PAR1 dimer via the CM sequence, and this PAR1directed CagA dimer would form a stable complex with SHP-2 through its two SH2 domains.
Because of the critical role of CagA in gastric carcinogenesis (7, 16 -19), it is important to elucidate the molecular basis underlying the morphogenetic activity of CagA. In this study, we investigated the role of PAR1 isoforms in induction of the hummingbird phenotype by CagA, and we obtained evidence that CagA-mediated inhibition of PAR1 kinases contributes to the development of the morphological change by perturbing microtubules and non-muscle myosin II.

EXPERIMENTAL PROCEDURES
Expression Vectors-pSP65SR␣-derived mammalian expression vectors for hemagglutinin (HA)-tagged East Asian CagA (CagA-ABD) and its phosphorylation-resistant mutant (CagAabd) were described previously (29). CagA-ABD-⌬CM was generated from CagA-ABD by deleting the 16-amino acid CM sequence by site-directed mutagenesis as described previously (43). A pEF-His-A-derived mammalian expression vector for T7 epitope-tagged human PAR1b/MARK2 was kindly provided by Dr. Shigeo Ohno (Yokohama City University, Japan). The cDNAs encoding human PAR1a/MARK3, PAR1c/MARK1, and PAR1d/MARK4 were purchased from OriGene Technologies. After adding a T7 epitope tag at the N terminus, the PAR1 cDNAs were inserted into pEF-His-A mammalian expression vector. PAR1b-K49R, PAR1b-S257I, and PAR1c-I275S were made by site-directed mutagenesis. Human Tau cDNA was tagged with FLAG epitope and inserted into the pcDNA3.1 mammalian expression vector.
Recombinant Adenoviruses-pAD/CMV/V5-DEST-derived adenovirus vectors for HA-tagged CagA-ABD and T7-tagged PAR1 were constructed using the ViraPower adenoviral expression system (Invitrogen) according to the manufacturer's instruction. A recombinant adenovirus transducing ␤-galactosidase was used as a negative control. Recombinant adenoviruses were amplified in human HEK293A cells, and the virus titers were determined by counting the plaques of HEK293A cells.
Cells, Transfection, and Transduction-Monkey COS-7 cells and Madin-Darby canine kidney (MDCK) II cells were cultured A, schematic view of T7-tagged human PAR1/MARK isoforms. The PAR1/ MARK isoforms include six domains. The N-terminal domain (N) is followed by the catalytic domain. Adjacent are the common docking sites responsible for targeting (T) to cell membrane and the ubiquitin-associated domain (UBA). A spacer domain connects the ubiquitin-associated domain with the C-terminal tail domain (C-tail). B, effect of CagA on the kinase activity of PAR1 isoforms. T7-PAR1 isoforms and FLAG-Tau were co-transfected into AGS cells with or without CagA-ABD-HA. Upper panel, level of Tau phosphorylation at Ser-262 was determined by immunoblotting (IB) with a specific anti-S262-Tau antibody. Lower panel, ratio of phosphorylated Tau to total Tau (anti-FLAG) was present in the histogram, in which the wild-type (WT) PAR1b was set as standard and arbitrary as 1. ␤-Actin was used as a protein loading control.
Antibodies-Anti-HA monoclonal antibody (3F10, Roche Applied Science) and anti-T7 polyclonal antibody (M-21, Santa Cruz Biotechnology) were used for immunoprecipitation of HA-tagged CagA and T7-tagged PAR1, respectively. Anti-␤-actin (C-11, Santa Cruz Biotechnology) was used to detect ␤-actin as a control. Anti-PAR1b antibody was a kind gift from Dr. Shigeo Ohno (Yokohama City University, Japan). Anti-S262-Tau antibody (Calbiochem catalog number 577814) and anti-FLAG antibody (M2, Sigma) were used to detect phosphorylated Tau and total Tau, respectively, on immunoblotting. Horseradish peroxidase-linked anti-mouse, anti-rat, and anti-rabbit IgG were purchased from GE Healthcare. Horseradish peroxidase-linked donkey anti-goat IgG was purchased from Santa Left panel, a constant amount of HA-tagged CagA (CagA-ABD-HA) was mixed in vitro with a constant amount of T7-tagged PAR1 isoform (T7-PAR1). The protein mixtures were incubated at 4°C for 2 h and were then immunoprecipitated with an anti-HA antibody. The anti-HA immunoprecipitates (IP) and total cell lysates (TCL) were resolved by SDS-PAGE and immunoblotted (IB) with the indicated antibodies. Right panel, relative CagA binding activity of each PAR1 isoform was determined by the ratio of PAR1 bound to CagA, which was calculated from the intensities of anti-T7 and anti-HA immunoblotting data measured by a lumino-image analyzer. Data obtained from five independent experiments are shown in the histogram. Error bars indicate mean Ϯ S.E., n ϭ 5. B, residues 250 -268 of the kinase catalytic domain of PAR1b, which are essentially involved in CagA binding, were aligned with those of corresponding regions in other PAR1 isoforms. Identical residues are shown in a box and % identities to the PAR1b sequence are shown to the right. C, in vitro binding assay of CagA with PAR1b mutant. Upper panel, a constant amount of HA-tagged CagA (CagA-ABD-HA) was mixed in vitro with a constant amount of T7-tagged wild-type (WT) PAR1b or PAR1b-S257I. Lower panel, amount of PAR1b-S257I that bound to CagA was quantitated and compared with that of wild-type PAR1b. The CagA binding activity of wild-type PAR1b was taken as 1. Error bars indicate mean Ϯ S.E., n ϭ 3. D, in vitro binding assay of CagA with PAR1c mutant. Upper panel, a constant amount of HA-tagged CagA (CagA-ABD-HA) was in vitro mixed with a constant amount of T7-tagged wild-type (WT) PAR1c or PAR1c-I275S. Lower panel, amount of PAR1c-I275S that bound to CagA was quantitated and compared with that of wild-type PAR1c. The CagA binding activity of wild-type PAR1c was taken as 1. Error bars indicate mean Ϯ S.E., n ϭ 3.
In Vitro Protein Interaction Assay-In vitro protein interaction assay was performed following the procedure described previously (43). In brief, COS-7 cells were singly transfected with the expression vector for HA-tagged CagA-ABD or T7-tagged PAR1 isoform. Cells were harvested at 36 h after transfection and lysed in lysis buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 5 mM EDTA, 1% Brij-35) containing 2 mM Na 3 VO 4 , 2 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 10 g/ml trypsin inhibitor, and 10 g/ml aprotinin (16). A constant amount of HA-tagged CagA-ABD expressed in COS-7 cells was mixed with a constant amount of each of the T7-PAR1 isoforms expressed in COS-7 cells. Total amount of proteins and volume were adjusted to the same level using lysates prepared from COS-7 cells transfected with a control vector. Mixed cell lysates were incubated at 4°C for 2 h before performing immunoprecipitation.
Immunoprecipitation and Immunoblotting-Total cell lysates and anti-HA immunoprecipitates were subjected to SDS-PAGE. Separated proteins were transferred to polyvinylidene difluoride membrane filters (Millipore), and the filters were incubated with a primary antibody (1:1000, 90 min at room temperature) and then with a secondary antibody (1:10,000, 45 min at room temperature). Proteins were visualized by using Western blot chemiluminescence reagent (PerkinElmer Life Sciences). Intensities of chemiluminescence on the immunoblotted filter were quantitated using a lumines-cence image analyzer (LAS-1000, Fujifilm) as described previously (29,31).
Transepithelial Electrical Resistance (TER)-MDCK II cells (1.5 ϫ 10 5 ) were seeded on a Transwell filter and cultured for 3 days to make a polarized monolayer. Polarized MDCK II cells were then transduced with adenovirus at the m.o.i. of 240:1 for each adenovirus. TER was measured every 24 h for 4 days using a Millicell-ESR epithelial voltohmmeter (Millipore). The specific TER value was calculated by subtracting the background value of the filter and the medium obtained from blank wells that had been cultured in parallel.
Quantitative Analysis of Focal Adhesions-AGS cells were fixed and stained with an anti-paxillin antibody to visualize focal adhesions. Pictures were acquired on a fluorescence microscope on an 8-bit scale. Focal adhesions were quantified using ImageJ software (NIH, rsbweb.nih.gov). In brief, images obtained from the anti-paxillin staining were converted to binary, and objects with a size greater than 0.5 m 2 were counted as focal adhesions. All images were background-subtracted prior to measurements. Nonspecific staining was removed manually.
Statistics-Statistical analyses for induction of the hummingbird phenotype by CagA in AGS cells were performed by Student's t test. The effect of blebbistatin on the AGS cells were analyzed by 2 test. Error bars represent mean Ϯ S.E. p Ͻ 0.05 was considered to be statistically significant.

RESULTS
CagA Is a Universal Inhibitor of PAR1/MARK Isoforms-In a previous study, we demonstrated that CagA specifically binds to the 27-amino acid stretch that is present in the C-terminal region of the PAR1b kinase domain (19). Interaction of CagA with PAR1b inhibits the kinase activity both in vitro and in vivo. Because CagA also interacts with other PAR1 isoforms via the CM sequence (19), it is a logical conclusion that CagA acts as a universal inhibitor of PAR1 kinases. To verify this, we generated pEF-based mammalian expression vectors for all of the human PAR1 isoforms (Fig. 1A). We also generated a kinasedead PAR1b mutant, PAR1b-K49R, in which the lysine residue on the ATP-binding site in the kinase domain of PAR1b was replaced by an arginine residue (35,42). We then expressed each of PAR isoforms together with the PAR1 kinase substrate, the Tau protein, in AGS human gastric epithelial cells in the presence or absence of CagA. Whereas the kinase-dead PAR1b-K49R failed to phosphorylate Tau, all the PAR1 isoforms induced Tau phosphorylation, which was inhibited by co-expression of CagA (Fig. 1B).
Differential CagA Binding Activities among PAR1/MARK Isoforms-To investigate the interaction of CagA with PAR1 isoforms in more detail, HA-tagged CagA or T7-tagged PAR1 isoform was singly expressed in COS-7 cells. The COS-7 cell lysates containing CagA was then mixed with those containing one of the PAR1 isoforms, and the mixed cell lysates were subject to an in vitro protein interaction assay (43). Among the four PAR1 isoforms, PAR1b displayed the strongest CagA binding activity, followed by PAR1d/MARK4 and PAR1a/ MARK3. PAR1c/MARK1 exhibited the weakest activity for binding to CagA ( Fig. 2A).
As noted above, the 27-amino acid CagA-binding sequence of PAR1b (residues 250 -276) is highly conserved among the PAR1 isoforms (19). In particular, comparison of residues 250 -268 of PAR1b, which have been reported to be indispensable for CagA interaction (43), with those of the corresponding regions in other PAR1 isoforms revealed that the identities of the amino acid sequences were 78 -84% (Fig. 2B). In the CagA-binding regions of PAR1b (the strongest CagA interactor) and PAR1c (the weakest CagA interactor), Ser-257 of PAR1b and Ile-275 of PAR1c were physicochemically the most diverged amino acid residues. To determine whether the differences in CagA binding activity among PAR1 isoforms were because of the amino acid variations within the CagA-binding region, the PAR1b-S257I mutant in which Ser-257 of PAR1b was replaced by Ile and the PAR1c-I275S mutant in which Ile-275 of PAR1c was replaced by Ser were generated by site-directed mutagenesis. When compared with wide type PAR1, PAR1b-S257I showed a decreased activity for binding to CagA (Fig. 2C), whereas PAR1c-I275S showed an increased CagA binding activity (Fig.  2D). Thus, the difference in CagA binding activity among PAR1 isoforms is primarily attributable to the sequence variations in the CagA-binding regions.
Role of PAR1 Isoforms in CagA-induced Disruption of Tight Junctions-In polarized epithelial cells, CagA elicits junctional and polarity defects, which were rescued by co-expression of PAR1b (19). To determine whether PAR1 isoforms other than PAR1b can also rescue the junctional defects caused by CagA, we generated recombinant adenoviruses for CagA and the PAR1 isoforms. Polarized MDCK II cells were then co-infected with adenoviruses transducing CagA and one of the PAR1 isoforms. The expression levels of CagA and PAR1 isoforms were almost comparable in MDCK II cells infected by the recombi- nant adenoviruses (Fig. 3A). The strength of the tight junctions was then quantitatively evaluated by measuring TER of the polarized MDCK monolayer. As shown in Fig. 3B, all of the PAR1 isoforms, including PAR1c which bound CagA most weakly, were capable of counteracting the CagA activity to disrupt tight junctions to comparable levels. The results indicate that all of the human PAR1 family members are involved in the maintenance of tight junctions. In turn, CagA disrupts tight junctions by acting as a universal inhibitor of PAR1 kinases.
Involvement of PAR1 Isoforms in Induction of the Hummingbird Phenotype by CagA-Induction of the hummingbird phenotype by CagA is counteracted by elevated levels of PAR1b. This in turn indicates that CagA-PAR1b interaction and subsequent inhibition of PAR1b is important for induction of the hummingbird phenotype (7,16). To determine whether PAR1b is the only PAR1 isoform that can inhibit the hummingbird phenotype, AGS cells were co-transfected with expression vectors for HA-tagged CagA and one of the T7-tagged PAR1 isoforms. At 24 h post-transfection, cells were stained with anti-HA and anti-T7 antibodies, and the percentages of hum-mingbird cells in CagA single-positive or CagA/PAR1 double-positive cells were calculated. In the morphological analysis, hummingbird cells were defined as cells with long and thin protrusions, the length of which is more than 2-fold the width of the cell body. Compared with CagA single-positive cells, CagA/ PAR1 double-positive cells included a significantly decreased number of hummingbird cells, in which the length of the protrusion was also reduced (Fig. 4, A and B). In these PAR1-transfected cells, all of the PAR1 isoforms were capable of attenuating the CagA-induced hummingbird phenotype at equivalent levels (Fig. 4B), and the levels of CagA as well as PAR1 isoforms expressed were comparable (Fig.  4C). Hence, all of the PAR1 isoforms can counteract the morphogenetic activity of CagA. Thus, systemic inhibition of PAR1 isoforms by CagA must play a role in induction of the hummingbird phenotype.
Effect of CagA on Subcellular Distribution of PAR1 Isoforms-To elucidate the mechanism by which PAR1 inhibition contributes to the morphogenetic activity of CagA, we next investigated the intracellular distribution of PAR1 isoforms in the presence or absence of CagA. To do so, AGS cells were transfected with one of the T7-tagged PAR1, and subcellular distribution of the PAR1 isoforms was examined by using a confocal microscope. In AGS cells, all of the ectopically expressed PAR1 isoforms were diffusely distributed in the cytoplasm, as were the cases with Chinese hamster ovary (CHO) cells (35,37) and HEK293 cells (38) (Fig. 5A). When CagA was co-expressed with one of the PAR1 isoforms, however, a significant fraction of PAR1 was translocated from the cytoplasm to the plasma membrane (Fig. 5B). Because CagA specifically localizes to the plasma membrane, this result indicates that the membrane-tethered CagA recruited PAR1 isoforms from the cytoplasm to the plasma membrane, where CagA inhibited the PAR1 kinase activity.
Effects of PAR1 Isoforms on Microtubules and Microfilaments-In mammalian cells, all of the PAR1 isoforms function as MARKs that phosphorylate MAPs, such as MAP2, MAP4, and Tau, and thereby regulate the dynamics of the microtubules (35,37). Because cell morphology is determined by the cooperative action of microtubules and microfilaments (45,46), it is possible that inhibition of PAR1 by CagA promotes the development of the hummingbird phenotype through perturbing microtubules. To test this possibility, we examined the role of each PAR1 isoform on microtubules in gastric epithelial cells (Fig. 6A). In AGS cells that did not express ectopic PAR1, microtubules were radiated from the centrosome and reached the cell periphery upon staining with anti-␤-tubulin antibody (Fig. 6A, see cells that are negative for anti-T7 staining). In contrast, in cells ectopically expressing one of the PAR1 isoforms, microtubules lost the radial appearance and primarily surrounded the nucleus, which was concomitantly associated with a decrease in anti-␤-tubulin staining in the cell periphery (Fig.  6A, see cells that are positive for anti-T7 staining). Furthermore, ectopic PAR1-expressing cells became smaller and more rounded than nonexpressing cells. Consistent with results of a previous study using CHO cells (35,37), this observation indicated that PAR1 kinases destabilize microtubules of AGS cells. We also visualized F-actin filaments in AGS cells with rhodamine-phalloidin staining (Fig. 6B). In AGS cells that did not express ectopic PAR1, F-actins were primarily organized into cortical actin bundles anchored to focal adhesions (Fig. 6B, see cells that are negative for anti-T7 staining). In cells expressing one of the PAR1 isoforms, however, the cortical F-actin bundles were markedly reduced (Fig. 6B, see cells that are positive for anti-T7 staining). Thus, PAR1 kinases not only regulate the dynamics of microtubules but also affect the reorganization of actin cytoskeletal system in AGS cells.
Ablation of PAR1-mediated Microtubule Destabilization by CagA-The above-described observations indicated that CagA counteracts the ability of PAR1 to destabilize microtubules upon physical complex formation. Accordingly, we next investigated the effect of CagA on the PAR1-regulated microtubule dynamics. In AGS cells co-expressing CagA and PAR1b, destabilization of microtubules by PAR1b was abolished (Fig. 7A). In these cells, the majority of CagA and PAR1b colocalized to the membrane as shown in Fig. 5B. AGS cells were next transfected with the T7-tagged PAR1b and HA-tagged CagA mutant, CagA-ABD-⌬CM, that specifically lacks the PAR1-binding CM sequence (Fig. 7B) (43). Although the CagA mutant was still capable of localizing to the plasma membrane, it neither recruited PAR1b to the cell membrane nor abolished the ability of PAR1 to destabilize the microtubules (Fig. 7B). Furthermore, CagA-ABD-⌬CM failed to induce the hummingbird phenotype as reported previously (43). These observations support the conclusion that CagA-mediated PAR1 inhibition and subsequent perturbation of microtubules are critically involved in the morphogenetic activity of CagA. To verify the above-described conclusion, we then knocked down PAR1b expression using siRNA in AGS cells (Fig. 8A). PAR1b knockdown per se induced morphological change in AGS cells, which was characterized by the appearance of multiple short protrusions (Fig. 8B). Expression of CagA in the PAR1b-knockdown cells elicited a dramatic increase in the number of hummingbird cells (Fig. 8, B and C). The observation consolidates that inhibition of PAR1 activity is critical in induction of the hummingbird phenotype by CagA. The conclusion was further supported by the experiment showing that, in contrast to the case of wild-type PAR1b, ectopic expression of a kinase-dead PAR1b, PAR1b-K49R, failed to inhibit induction of the hummingbird phenotype by CagA. Rather, PAR1b-K49R potentiated the hummingbird phenotype, probably due to inhibition of endogenous PAR1 by PAR1b-K49R, which may act as a dominant-negative PAR1 (Fig. 8D).
Involvement of Microtubules in Induction of the Hummingbird Phenotype by CagA-The hummingbird phenotype involves malfunctioning of focal adhesions, which is caused at least partly by the inhibition of FAK through CagA-deregulated SHP-2 (31). Because microtubules play an important role in the focal adhesion turnover (47)(48)(49)(50)(51), CagA-PAR1 interaction and subsequent inhibition of PAR1 kinase activity might also contribute to the hummingbird phenotype by perturbing focal adhesions. To test this possibility, we investigated the effect of a phosphorylation-resistant CagA (CagA-abd), which binds PAR1 but not SHP-2 (29), on the formation of focal adhesions. Like wild-type CagA (CagA-ABD), expression of phosphorylation-resistant CagA (CagA-abd) caused a decrease in the number of focal adhesions (Fig. 9, A and B, compare the numbers of focal adhesion spots (green dots) with CagA-negative (red) and CagA-positive (blue) cells in A). Wild-type CagA also reduced the sizes of focal adhesion spots, whereas phosphorylation-resistant CagA failed to alter the sizes of the focal adhesion spots (Fig. 9, A and C, compare the sizes of focal adhesion spots (green dots) with CagA-ABD-positive cells (blue cells in left panels) and CagA-abdpositive cells (blue cells in right panels) in A). Hence, the effect of phosphorylation-resistant CagA on focal adhesions was attenuated compared with the effect of wild-type CagA. This observation is consistent with the notion that wild-type CagA inhibits the formation of focal adhesions by both activating SHP-2 and inhibiting PAR1, whereas phosphorylation-resistant CagA inhibits the formation of focal adhesions only through inhibition of PAR1. Again, the findings support the idea that systemic inhibition of PAR1 family kinases by CagA, which results in stabilization of microtubules, also contributes to the hummingbird phenotype by subverting focal adhesions.

Involvement of Non-muscle Myosin II in the Hummingbird
Phenotype Induced by CagA-The thin and long protrusions of the hummingbird phenotype are mechanistically due to retraction failure at the rear end of the moving cell (52). It has been reported that non-muscle myosin II, especially myosin IIB, is responsible for retraction of the rear tail (53)(54)(55). The fact that myosin II is physically and functionally associated with PAR1 (56) raises the possibility that inhibition of PAR1 by CagA may also be involved in the hummingbird phenotype through perturbing myosin II functions. To test this possibility, we investigated the effect of a specific myosin II inhibitor, blebbistatin (57), on induction of the hummingbird phenotype in AGS cells. Blebbistatin inhibits both the ATPase and gliding motility activities of vertebrate non-muscle myosin II without inhibiting myosin light chain kinase (MLCK). It rapidly disrupts the directed cell migration without affecting the assembly of microtubules (57). Treatment of AGS cells with 10 M blebbistatin for 5 h induced short protrusions of cells (Fig. 10A), which are similar to those observed in the PAR1b-knockdown cells (Fig.  8B). Given this, we next transfected AGS cells with a CagA expression vector and treated cells with a suboptimal concentration of blebbistatin (1 M) for 12 h, which on its own does not induce any morphological changes in AGS cells, before staining. The results of the experiment revealed that blebbistatin treatment significantly increased the number of cells with the hummingbird phenotype by CagA (Fig. 10, B and C). The length of protrusion in the hummingbird cell was also increased in the presence of blebbistatin (Fig. 10, B, D, and E). Thus, induction of the hummingbird phenotype by CagA was both quantitatively and qualitatively potentiated by inhibition of non-muscle myosin II. Although CagA-mediated SHP-2 deregulation and microtubule stabilization are essential for induction of the hummingbird phenotype (16, 58), perturbation of PAR1-regu-

DISCUSSION
In this study, we found that H. pylori CagA interacts with the PAR1 kinases and inhibits the kinase activity, which in turn causes microtubule stabilization. Excessively stabilized microtubules then promote disassembly of focal adhesions, contributing to the reduced cell-substrate interaction and subsequent induction of the hummingbird phenotype. We also found that malfunctioning of non-muscle myosin II is involved in the rear retraction defect, a characteristic of the hummingbird phenotype.
This study extends our previous work by showing that both the bacterial protein CagA acts as a universal inhibitor of PAR1 kinases and CagA interacts with all four PAR1 isoforms, in order of strength as follows: PAR1b Ͼ PAR1d Ն PAR1a Ͼ PAR1c. The PAR1 isoforms are widely expressed in a variety of cell types and are thought to share redundant biological functions (35,37). Indeed, par1b-null mice are viable and do not show overt polarity defects, supporting functional compensation among the PAR1 isoforms (59,60). The observation that the CagA-mediated junctional defect is abolished by any member of the PAR1 family kinases also supports the shared role among the PAR1 isoforms in the regulation of cell polarity.
AGS gastric epithelial cells display a polygonal morphology with cortical distribution of filamentous actin fibers (F-actins). Consistent with the results of previous studies (35)(36)(37), AGS cells ectopically expressing one of the PAR1 isoforms became smaller and more rounded than nonexpressing cells. In these cells, PAR1 isoforms were distributed diffusely in the cytoplasm and induced destabilization of microtubules as described previously in CHO cells. It is also notable that ectopically expressed PAR1 decreased cortical F-actins, and this was concomitantly associated with a slight increase in cytoplasmic F-actins. Thus, all of the PAR1 isoforms are capable of destabilizing the microtubule and actin cytoskeleton systems in a redundant fashion. The actin cytoskeletal changes caused by PAR1 may be due at least in part to the functional interaction between microtubules and a RhoA-specific guanine nucleotide exchange factor such as GEF-H1 (61)(62)(63). GEF-H1 activity is suppressed when it binds to microtubules and is increased when it is released from microtubules (64). Accordingly, disruption of microtubules by PAR1 causes an increase in activity of GEF-H1, which in turn activates RhoA and thereby causes actin-cy- toskeletal reorganization and myosin II-dependent contraction, which lead to cell rounding.
Deregulation of SHP-2 by CagA is critically involved in the hummingbird phenotype (16,17,29,31). We previously reported that CagA-activated SHP-2 dephosphorylates FAK at its activating phosphorylation residues (Tyr-375, Tyr-575, and Tyr-577), which results in FAK inactivation (31). Down-regulation of FAK then impairs turnover of the focal adhesion, the site of clustered integrins and associated molecules that are linked to actin stress fibers (65)(66)(67), thereby promoting induction of the hummingbird phenotype. This study revealed that induction of the hummingbird phenotype requires systemic inhibition of PAR1 kinases in addition to deregulated SHP-2 (19,43). How then is PAR1 involved in the hummingbird phenotype? Recent studies have shown that microtubules make direct contact with focal adhesions and promote their disassembly by delivering dynamin and ACF7, which trigger endocytosis of focal adhesion components such as integrins (51,64). Thus, interaction of CagA with PAR1 isoforms impairs the ability of PAR1 to destabilize microtubules and thereby promotes disassembly of focal adhesions.
Mechanistically, the hummingbird phenotype is caused by a defect of rear retraction during cell movement (52). This morphological abnormality is reminiscent of that observed in cells lacking non-muscle myosin II (53,54). Indeed, myosin II-mediated contraction is thought to generate force to retract the rears of migrating cells and to maintain a polarized cell shape (63,64). We found in this study that inhibition of nonmuscle myosin II potentiates induction of the hummingbird phenotype by CagA. Notably, Guo and Kemphues (56) reported a physical and functional interaction between PAR1 and nonmuscle myosin II in establishing cell polarity in Caenorhabditis elegans zygotes, raising the possibility that PAR1 regulates myosin II functions through complex formation and/or phosphorylation and that CagA stimulates the formation of long protrusions, the hallmark of the hummingbird phenotype, by perturbing non-muscle myosin II via PAR1 inhibition.
This study revealed that microtubules and non-muscle myosin II are key structural elements underlying the morphogenetic activity of CagA. First, CagA subverts focal adhesions by inhibiting FAK through CagA-mediated SHP-2 activation and by stabilizing microtubules through CagA-mediated PAR1 inhibi- tion. Second, CagA perturbs non-muscle myosin II and thereby prevents retraction of rear protrusions during cell movement. Depolarization of epithelial cells is associated with the conversion of apical-basolateral polarity to antero-posterior (front-back) polarity. The antero-posterior polarity, which is also regulated by PAR1 kinases, allows cells to migrate in a directional fashion (68). Focal adhesion assembly at the anterior front and dissolution at the posterior end are critical for migration in many adherent cells (55,66,69). Hence, the hummingbird phenotype may need to be understood in terms of the antero-posterior polarity defect caused by the CagA-PAR1 interaction.
Strong association of H. pylori cagA-positive strains with the development of gastric carcinoma indicates the importance of CagA in gastric carcinogenesis. Indeed, the oncogenic potential of CagA has recently been demonstrated by studies using cagA-transgenic mice (15). It has long been thought that the CagA-induced hummingbird phenotype is related to the oncogenic activity of CagA. Pathophysiological relevance for the hummingbird phenotype in gastric carcinogenesis has also been revealed by recent clinical observations that infection with H. pylori carrying CagA with stronger ability to induce the hummingbird phenotype is more closely associated with gastric carcinoma (20 -23). The hummingbird phenotype requires CagA-mediated perturbation of SHP-2 and PAR1, deregulation of which are substantially involved in the oncogenic process. SHP-2 is a bona fide oncoprotein, gain-offunction mutations of which are associated with various human malignancies (28). Also, PAR1 plays an essential role in the development and maintenance of epithelial cell polarity, disruption of which is a hallmark of carcinoma. PAR1d have been suggested to be tumor suppressors of pancreatic carcinoma (70). Furthermore, inactivation of LKB1, an upstream activating kinase of PAR1, causes Peutz-Jeghers syndrome, which predisposes to the development of gastrointestinal tumors (71). Thus, molecular mechanisms underlying induction of the hummingbird phenotype by CagA may also be critically involved in in vivo tumorigenesis by CagA. Because hummingbird cells display elevated motility with reduced cell-cell and cell-substrate interactions, the morphogenetic activity of CagA may further contribute to invasion and metastasis of gastric carcinoma cells.