Functional analysis of Shigella VirG domains essential for interaction with vinculin and actin-based motility.

The VirG (IcsA) protein of Shigella is required for recruitment of host actin filament (F-actin) by intracellularly motile bacteria. An N-terminal 80-kDa VirG portion (alpha-domain) is exposed on the bacterial surface, while the following C-terminal 37-kDa portion (beta-core) is embedded in the outer membrane. Here, we report that the surface exposed alpha-domain of VirG possesses two distinct functional domains; one is the N-terminal two-thirds portion of the alpha-domain which is required for eliciting F-actin assembly on the bacteria in infected cells, and the other one is the rest of the C-terminal portion of the VirG alpha-domain, which is essential for the asymmetric distribution of VirG on the bacterial surface. Furthermore, we found that vinculin, an actin-binding cytoskeletal protein, accumulates on the surface of bacteria expressing VirG in infected cells, and that the distribution of vinculin coincided with the distribution of VirG and assembled F-actin. The vinculin accumulation depended on the expression of the alpha-domain VirG portion required for F-actin assembly, but the recruitment of vinculin on Shigella appeared prior to the appearance of F-actin in the infected cells. Analysis of proteins interacting with VirG using Xenopus laevis eggs extracts revealed that vinculin was a protein that bound to the alpha-domain portion. This was further confirmed using purified chicken gizzard vinculin, in that the 95-kDa vinculin head part, but not the 30-kDa tail part, directly bound to the alpha-domain portion. These results suggest a possible role for vinculin in recruitment of F-actin to the VirG moiety exposed on Shigella in infected mammalian cells.

The VirG (IcsA) protein of Shigella is required for recruitment of host actin filament (F-actin) by intracellularly motile bacteria. An N-terminal 80-kDa VirG portion (␣-domain) is exposed on the bacterial surface, while the following C-terminal 37-kDa portion (␤-core) is embedded in the outer membrane. Here, we report that the surface exposed ␣-domain of VirG possesses two distinct functional domains; one is the N-terminal twothirds portion of the ␣-domain which is required for eliciting F-actin assembly on the bacteria in infected cells, and the other one is the rest of the C-terminal portion of the VirG ␣-domain, which is essential for the asymmetric distribution of VirG on the bacterial surface. Furthermore, we found that vinculin, an actinbinding cytoskeletal protein, accumulates on the surface of bacteria expressing VirG in infected cells, and that the distribution of vinculin coincided with the distribution of VirG and assembled F-actin. The vinculin accumulation depended on the expression of the ␣-domain VirG portion required for F-actin assembly, but the recruitment of vinculin on Shigella appeared prior to the appearance of F-actin in the infected cells. Analysis of proteins interacting with VirG using Xenopus laevis eggs extracts revealed that vinculin was a protein that bound to the ␣-domain portion. This was further confirmed using purified chicken gizzard vinculin, in that the 95-kDa vinculin head part, but not the 30-kDa tail part, directly bound to the ␣-domain portion. These results suggest a possible role for vinculin in recruitment of F-actin to the VirG moiety exposed on Shigella in infected mammalian cells.
The ability of Shigella to enter colonic epithelial cells and to subsequently spread within and between the cells is a prerequisite for causing dysentery. The capacity of the bacteria to spread in the cytoplasm and then move into adjacent epithelial cells is known as intra/intercellular spreading, respectively, and requires the bacterial functions encoded by the virG (icsA) gene (1)(2)(3).
Shigella enter into mammalian cells by inducing bacterial directed endocytosis (4 -6), lysing the surrounded endocytic vacuole, and then multiplying within the cytoplasm, in which the bacteria polymerize actin or/and recruit filamentous actin (F-actin) on their surfaces. The F-actin clot then rearranges into a long tail which remains stationary in the cytoplasm and is left behind by bacteria moving ahead (2,7,8). When the spreading bacterium makes contact with the inner surface of the host cell plasma membrane, a long membrane protrusion with the F-actin tail behind it develops and is endocytosed by the adjacent cell, resulting in the surrounding of the bacterium by a double membranous vacuole (9 -11). After disruption of the vacuole, bacteria are released into the new host cell cytoplasm and multiply again. Thus, actin-based bacterial movement is essential for continuing intercellular spreading among colonic epithelial cells.
VirG is a surface-exposed outer membrane protein, and the sequence of the virG gene reveals that it is a 1102-amino acid protein with an atypical signal peptide (3,8,12). Subsequent studies have indicated that the 116-kDa VirG protein consists of three distinctive domains, the N-terminal signal sequence (amino acids 1-52), the 706 amino acids ␣-domain (amino acids 53-758), and the 344 amino acids C-terminal ␤-core (amino acids 759-1102) (12). The ␣-domain is presented on the surface of Shigella and contains repetitive sequences rich in Gly residues (3,8,(12)(13)(14), while the ␤-core is embedded in the outer membrane, possibly forming a ␤-barrel structure, through which the ␣-domain is translocated across the outer membrane (12). Although the ␣-domain portion is cleaved off from the ␤-core and released into the bacterial environment under in vitro and in vivo conditions (8,13,14), it cannot interact with the F-actin tail in infected cells (14,15). Nevertheless, the assembly of F-actin near the surface of intracellular Shigella is absolutely dependent on the surface presentation of VirG ␣-domain and the formation of an actin tail depends on the asymmetric distribution of VirG (8,12,16,17). Recently, to demonstrate that VirG is the sole bacteria factor needed for movement in infected mammalian cells, experiments on actinbased movement were carried out using Xenopus laevis eggs extracts (18,19). In these studies, heterologous expression was achieved by exploiting an OmpT Ϫ (OmpT is a trypsin-like outer membrane protease) VirG-expressing strain of Escherichia coli K12, in which the cleavage of VirG and thus, the release of the ␣-domain into the bacterial environment was prevented (13). This strain displayed an asymmetric distribution of VirG on the bacterial body and was capable of intra/intercellular movement (19). In addition, VirG was only found on the bacterial surface required for actin-based movement.
In this context, we constructed various internal in-frame deletions of the ␣-domain and attempted to delineate the functional domain required for asymmetric distribution of VirG on the bacterial surface or for the actin-based motility of bacteria in infected cells. Our data indicate that bacteria expressing VirG variants lacking the N-terminal two-thirds of the ␣-domain cannot aggregate F-actin or form an actin-tail, while the bacteria expressing VirG variant lacking the remaining Cterminal portion of the ␣-domain fail to display an asymmetric distribution of VirG on the outer surface. In this study, we found that accumulation of vinculin, an actin-binding protein involved in formation of focal adhesions (21), was elicited by bacteria expressing VirG in infected cells, and that distribution coincided with the localized accumulation of VirG and F-actin. However, the vinculin clot appears faster than the F-actin assembly on the bacterial surface. Subsequent in vitro studies using X. laevis eggs extracts or vinculin from chicken gizzard indicated that vinculin is a protein interacting with VirG, and that the purified vinculin head part can directly bind to the ␣-domain portion required for the F-actin aggregation. Based on this evidence, we speculate that vinculin plays a role in recruitment of F-actin by VirG expressed on intracellular Shigella.

EXPERIMENTAL PROCEDURES
Construction of In-frame Deletion Mutants in VirG-Plasmid pD10-1 was a derivative of pD10 (a virG clone) (3), in which a DNA sequence between the EcoRI site and nucleotide 193 upstream from the 5Ј end of the virG gene on pD10 was deleted using exonuclease III. The ScaI site on this plasmid was then replaced with a BamHI linker and the resultant plasmid named pD10-1. pD10-1 was subsequently used as a template for polymerase chain reaction (PCR) 1 in construction of in-frame deletions. A DNA fragment encompassing nucleotides from position 957 downstream from the 5Ј end through to nucleotide 1239 was amplified by PCR using primers 5Ј-GCTCTAGACAAAATGTAGCAGGTA-ATGC-3Ј and 5Ј-CCCAAGCTTCCAGACTACTGATTCCAGC-3Ј containing XbaI and HindIII sites in the 5Ј tail, respectively. The 278-bp XbaI-HindIII segment was then replaced with the XbaI-HindIII segment on pD10-1; the resulting plasmid encoding a VirG variant lacking amino acids 104 -319 was designated as pD10-1virG1. A DNA fragment encompassing nucleotides from position 303 downstream from the 5Ј end through to nucleotide 956 was amplified by PCR using primers 5Ј-GCTCTAGAACTAAGCTACGGATT-3Ј and 5Ј-AAAAGTACTTGTTC-CATCATCTTCTTTACC-3Ј containing XbaI and ScaI sites in the 5Ј tail, respectively. The 660-bp XbaI-ScaI segment was replaced with the XbaI-ScaI segment on pD10-1; the resulting plasmid encoding a VirG variant lacking amino acids 320 -506 was designated as pD10-1virG2. A DNA fragment encompassing nucleotides from position 2187 downstream from the 5Ј end through to nucleotide 3110 was amplified by PCR using primers 5Ј-AAAAGTACTACCAATAAGTGGTATC-3Ј and 5Ј-CCGGAGCTCCATTTCCAATTC-3Ј containing ScaI and SacI sites in the 5Ј tail, respectively. The 935-bp ScaI-SacI segment was replaced with the ScaI-SacI segment on pD10-1; the resulting plasmid encoding a VirG variant lacking amino acids 509 -729 was designated as pD10-1virG3.
A DNA fragment encompassing nucleotides from position 1524 downstream from the 5Ј end through to nucleotide 2633 was amplified by PCR using primers 5Ј-GCTCTAGAATTCTGGCAGATAATCTC-3Ј and 5Ј-GCTATCCTGATATCCATAAGC-3Ј containing XbaI and EcoRV sites, respectively. The 1108-bp XbaI-EcoRV segment was replaced with the XbaI-EcoRV segment on pD10-1; the resulting plasmid encoding a VirG variant lacking amino acids 104 -506 was designated as pD10-1virG4. All the nucleotide sequences (ϳ100 nucleotides) encompassing each of the deletion ends of the pD10-1 derivatives were verified by DNA sequencing.
Immunofluorescence of in Vitro Grown Bacteria-Immunolabeling of bacteria was performed as described (23) using rabbit VRG-N2 antibody and fluorescein-isothiocyanate (FITC)-conjugated anti-rabbit IgG (Sigma). Preparations were observed with a confocal laser-scanning microscope MRC-1024 (Bio-Rad). The cut-off value was raised to 120 to estimate the distribution of the highest fluorescence intensity.
Cell Culture, Infection, and Triple Fluorescence Labeling of VirG, Vinculin, and F-actin-Human 293 fibroblasts (ATCC CRL1573) were cultured in Dulbecco's modified Eagle's medium with 25 mM HEPES and 10% fetal calf serum. Immunofluorescence of Shigella flexneriinfected cells was performed as described (24). VirG expressed on the bacterial surface was labeled with rabbit VRG-N2 antibody and Cy5conjugated anti-rabbit IgG (Amersham, Buckinghamshire, United Kingdom). Vinculin was labeled with anti-human vinculin monoclonal antibody hVIN-1 and FITC-conjugated anti-mouse IgG (Sigma). F-actin was visualized using rhodamin-phalloidin (Molecular Probes Inc., Eugene, OR). Accumulation of vinculin or F-actin around the intracellular bacteria was defined as areas where the intensity was higher than an arbitrary threshold of 80 estimated in noninfected cells.
In Vitro Binding Assay of X. laevis Eggs Extracts with GST Fusion Proteins-Meiotically arrested cytoplasmic extracts of X. laevis eggs were prepared according to Theriot et al. (26), and preincubated with glutathione-Sepharose 4B (Pharmacia) equilibrated in XB (100 mM KCl, 1 mM MgCl 2 , 0.1 mM CaCl 2 , 10 mM HEPES, pH 7.7, 50 mM sucrose) for 30 min at room temperature. GST-␣-domain fusion proteins (20 g) immobilized to glutathione-Sepharose were washed in XB prior to use. Precleared 100 l of extracts were added to 30 l of packed beads and incubated for 3 h at room temperature. Beads were recovered by centrifugation at 5,000 ϫ g in a microtube centrifuge and washed twice in 1 ml of XB. Bound proteins were solublized in SDS sample buffer (2% SDS, 4% 2-mercaptoethanol, 10% glycerol, 0.1 M Tris-HCl, pH 6.8). The eluted proteins were then subjected to SDS-polyacrylamide gel electrophoresis (PAGE) and immunoblotting as described before (14).
Purification of Vinculin and V8 Fragments-Chicken gizzard vinculin was prepared according to Feramisco and Burridge (27). During purification of the vinculin, EGTA (1 mM) was added to all buffers to inhibit calcium-activated proteases. To prepare the 95-and 30-kDa fragments, vinculin (ϳ1 mg/ml) was digested with Staphylococcus aureus V8 protease attached to cross-linked beaded agarose (Sigma) according to Johnson and Craig (28). The 95-kDa head and 30-kDa tail fragments were purified according to Groesch and Otto (29). Purity of the head and tail were determined by SDS-PAGE and Coomassie Blue 1 The abbreviations used are: PCR, polymerase chain reaction; bp, base pair(s); GST, glutathione S-transferase; FITC, fluorescein-isothiocyanate; PAGE, polyacrylamide gel electrophoresis. 2 T. Suzuki, unpublished results.

Co-precipitaion with GST-␣-domain Fusion Proteins on
Glutathione-Sepharose Beads-2.5 M GST fusion proteins were incubated for 3 h at room temperature with 0.5 M purified vinculin, the 95-kDa head portion or 30-kDa tail portion. Incubations (100 l) were performed in TEEAN, 0.5 mM 2-mercaptoethanol containing 1% bovine serum albumin. 50 l of a 50% slurry of glutathione-Sepharose in buffer was added and incubated for 30 min with rotating inversion. Beads were washed twice in 1 ml of buffer containing 0.1% bovine serum albumin, and solublized in SDS sample buffer before analysis.

Construction of Internal
In-frame Deletions of the VirG ␣-Domain-To delineate the VirG portion required for eliciting aggregation of F-actin and formation of an F-actin tail from Shigella in infected cells, we constructed internal in-frame deletions in the ␣-domain using pD10-1, a cloned virG plasmid (see "Experimental Procedures"). The resulting four derivatives were designated pD10-1virG1 through pD10-1virG4, lacking amino acids 104 -319 (pD10-1virG1), 320 -506 (pD10-1virG2), 509 -729 (pD10-1virG3), and 104 -506 (pD10-1virG4), respectively (Fig. 1). The respective VirG variant proteins produced from the plasmids were designated VirG1 through VirG4. All of the VirG variants were expressed under the control of the virG promoter and contained the VirG signal peptide sequences together with the ␤-core (Fig. 1). To see if each of the VirG variants were expressed from these plasmids, the plasmids were introduced into M94 (S. flexneri virG::Tn5) (1), and the whole cell protein extracts from each construct were analyzed by immunoblotting with VRG-N2 antibody, which was raised against a synthetic peptide corresponding to the region encompassing amino acids 82-100 of VirG (3). As determined by SDS-PAGE analysis, the VirG variants were all stably expressed, and their expected their sizes varied (VirG1, 96 kDa; VirG2, 97 kDa; VirG3, 93 kDa; and VirG4, 76 kDa) in M94 grown in BHI broth (data not shown).
Immunofluorescence Staining of in Vitro Grown S. flexneri Strains Expressing Various VirG Variants-To examine whether the lack of various ␣-domain portions affected the VirG expression on the bacterial surface, bacteria grown in BHI were immunostained with FITC-labeled VRG-N2 antibody and examined by confocal laser-scanning microscopy. As shown in Fig. 2, although individual bacteria expressed at slightly different levels, VirG1, VirG2, and VirG4 variants were all expressed on the bacterial surface in a polarized manner like the VirG wild type. Semiquantitative analysis of the distribution of the three VirG variants revealed deposition on the bacterial surface with a high density of the protein at one pole and a weak expression at the other pole (Fig. 2, A-C and E). In contrast, the distribution of VirG3 was less polarized, with a tendency to be distributed over the whole bacterial body (Fig.  2D). The distribution patterns of these VirG variants in infected cells such as 293 human fibroblasts (see below) or LLC-MK2 epithelial cells (data not shown) at 20 min after infection were essentially the same as those grown in vitro. Thus, these results suggested that the C-terminal part of the ␣-domain, corresponding to VirG amino acids 509 -729, was a portion of the VirG essential for the asymmetric distribution on the bacterial surface.
The N-terminal Two-thirds Portion of the ␣-Domain of VirG Is Required for F-actin Assembly by Intracellular Shigella-Bacteria expressing VirG (M94 carrying pD10-1) or the VirG variants (M94 carrying pD10-1virG1, pD10-1virG2, pD10-1virG3, or pD10-1virG4) were infected into 293 human fibroblasts, and after 80 min of infection the fibroblasts were immunostained and analyzed for the distribution of F-actin and VirG by confocal microscopy. The bacteria expressing wild-type VirG or VirG3 variant, but not the VirG1, VirG2, or VirG4 variants, were capable of eliciting F-actin assembly in the infected cells as examined by immunostaining with rhodaminlabeled phalloidin or Cy5-labeled VRG-N2 antibody. However, while the bacteria expressing wild-type VirG could form an F-actin tail (Fig. 3F), the bacteria expressing VirG3 failed to achieve this, and rather, became covered with F-actin (Fig. 3I). The bacteria expressing VirG variants were unable to move from one cell to another, because when MK2 epithelial cell monolayers were infected with these bacteria, none of them could form plaques even after 3 days incubation. These results support the premise that the unipolar distribution of VirG is a prerequisite for Shigella to elicit F-actin tail in infected cells (16,17), and suggest that the N-terminal two-thirds of the ␣-domain, which contains the eight Gly-rich repetitive sequences and the following ϳ80 amino acids is necessary for F-actin assembly (see Fig. 1).
Vinculin Accumulates at the Same Site as VirG Deposition on Shigella in Infected Cells-A previous study indicated that antibody to vinculin, a protein at focal adhesions interacting with actin and the focal adhesion proteins including talin (21,30), labeled the full-length of protrusions generated by moving S. flexneri in an infected cell as well as the bacteria at the protrusion tips (9). To investigate the role of vinculin in Shigella-induced F-actin assembly, 293 fibroblasts infected with bacteria expressing wild-type VirG (M94 carrying pD10-1) were stained after 20, 40, 60, and 80 min of infection with Cy5-labeled VRG-N2, FITC-labeled anti-human vinculin FIG. 1. Schematic representation of the structure of VirG and its variants. The three contiguous boxes represent the N-terminal signal sequence, ␣-domain, and ␤-core, respectively (12). Shaded areas indicate repetitive sequences rich in Gly residues (3). The ␤-core is indicated by cross-hatching. The bent lines correspond to the parts of VirG which are absent in the protein variants, while the numbers correspond to the positions of the amino acids in the wild-type VirG. Distribution of VirG and its variants on bacteria were based on the results in Fig. 2. Abilities of VirG and its variants to recruit F-actin assembly and/or vinculin accumulation on the intracellular bacteria were based on the results in Fig.  3. The focus-Plaque test measures bacterial ability to spread intercellularly (plaque ϭ able to spread) (11). The asterisk indicates that M94 carrying pD10-1virG3 elicited F-actin accumulation around the bacteria but could not form an F-actin tail. monoclonal antibody, or rhodamin-labeled phalloidin, and examined for the distributions of VirG, vinculin, and F-actin on the bacteria by confocal microscopy. After 20 min of infection, bacteria expressing VirG accumulated vinculin at the same sites as VirG deposition on the bacterium (Fig. 3, A and B). Approximately half (43 Ϯ 7%, n ϭ 119) of the vinculin-positive bacteria did not have a detectable F-actin clot as judged by immunostaining with rhodamin-labeled phalloidin (Fig. 3, B and C). The proportion of vinculin-positive but F-actin-negative bacteria to that of vinculin and F-actin-positive bacteria decreased as the infection period increased, and at 80 min after infection only ϳ10% (11 Ϯ 5%, n ϭ 142, Fig. 3, E and F) of vinculin-positive bacteria were actin-negative. Under the same conditions, infecting bacteria expressing VirG3 variant (M94 carrying pD10-virG3) displayed accumulation of vinculin over the whole bacterial body (Fig. 3H). Similarly, at 20 min postinfection ϳ50% (49 Ϯ 5%, n ϭ 138) of the vinculin-positive bacteria of M94 carrying pD10-1VirG3 did not have a detectable actin assembly on the bacterial surface, and at 80 min post-infection only ϳ10% (16 Ϯ 12%, n ϭ 114) of vinculinpositive bacteria lacking actin assembly. The other VirG variants, VirG1, VirG2, and VirG4, were unable to accumulate vinculin (data not shown). Thus, these results suggest that the bacteria expressing VirG elicit asymmetric vinculin accumula-tion in infected cells, which is followed by F-actin assembly, and that the N-terminal two-thirds of the ␣-domain, corresponding to amino acids 104 -506 is necessary for accumulation of vinculin as well as for assembly of F-actin.
Interaction between VirG and Vinculin in Xenopus Eggs Extracts-To investigate the possible interaction of vinculin with the ␣-domain of VirG, we used cytoplasmic extracts of X. laevis eggs, since this has been shown to support the actin-based motility of S. flexneri in a manner similar to that of infected mammalian cells (18,19). For this purpose, we constructed a plasmid encoding a GST-VirG ␣-domain fusion protein using PCR. The resulting plasmid (pGEX-␣1) encoded the ␣-domain (VirG amino acids 53-758) and an additional 21 amino acids of the N-terminal ␤-core (Fig. 4), and was affinity purified using a glutathione-Sepharose column. Subsequently, Xenopus egg extracts prepared according to the methods described by Theriot et al. (26) were incubated with glutathione-Sepharose ligated to GST-␣1, and the proteins bound to the GST-␣1 fusion protein were examined by SDS-PAGE. Although numerous proteins bound to the GST-␣1 fusion protein as detected by Coomassie Blue staining, immunoblotting with the anti-chicken vinculin monoclonal antibody revealed that the 120-kDa vinculin protein bound to GST-␣1 (Fig. 5B, lane 1) but not to GST itself (lane 7). Numerous proteins were also observed to bind to the GST protein itself using the same Xenopus egg extracts, indicating those were nonspecifically bound proteins. To further investigate whether or not the observed vinculin binding was specific for the ␣-domain portion which we had defined as the portion required for eliciting a vinculin clot in infected cells (see Fig. 1), a series of plasmids encoding various VirG ␣-domain portions were constructed in a similar manner to pGEX-␣1. The resulting plasmids were designated pGEX-␣2 to pGEX-␣6; pGEX-␣2 encoded the ␣-domain but lacked VirG amino acids 104 -319, pGEX-␣3 encoded VirG amino acids 53-506, pGEX-␣4 encoded VirG amino acids 53-319, pGEX-␣5 encoded VirG amino acids 320 -507, and pGEX-␣6 encoded VirG amino acids 53-101 (Fig. 4). The GST-␣-domain variants expressed from each of the plasmids were purified using a glutathione-Sepharose affinity column, then checked for size variation by SDS-PAGE (Fig. 5A) and immunoblotting with anti-GST antibody (see below). The Xenopus egg extracts were incubated with glutathione-Sepharose ligated to each of the GST-␣-domain variants, and the proteins bound were again examined in the same manner as above. As shown in Fig. 5B, the 120-kDa protein identified as vinculin bound to GST-␣1 and GST-␣3, but not to the other constructs, is in good agreement with our data showing that bacteria expressing wild-type VirG or VirG3 variant elicited vinculin accumulation in infected cells (Fig. 3).
Vinculin Head Binds to VirG-Vinculin has two principal domains, an N-terminal globular head and an elongated Cterminal tail (31). Recently, Johnson and Craig (32) identified an intramolecular interaction between the head and tail domains of vinculin, and found that the tail has an actin-binding site, while the head has an talin-binding site, but showed that the head-tail interaction normally masks these sites in the purified protein. Thus, to confirm the ability of vinculin to interact with VirG and further investigate its interactions with the head and tail part, we purified vinculin from chicken gizzard according to methods previously reported (27) and tested it for its ability to directly bind to the ␣-domain of VirG. After incubation of the purified vinculin together with each of the GST-␣-domain variants in a buffer for 3 h at room temperature, the GST-␣-domain variants were collected using glutathione-Sepharose beads and bound proteins analyzed by immunoblot- ting with anti-chicken vinculin antibody. In agreement with the above results, vinculin bound specifically to GST-␣1 and GST-␣3 but not to the other GST-␣-domain variants (Fig. 6A). Subsequently, we cleaved vinculin with V8 protease (28,29), and the resulting 95-kDa head and 30-kDa tail portions were incubated with GST-␣1, GST-␣3, and GST under the same conditions, and the bound vinculin portions examined by immunoblotting with anti-vinculin head or anti-vinculin tail monoclonal antibody. As shown in Fig. 6B, the vinculin head, but not the tail, bound to GST-␣1 and GST-␣3 (see "Discussion"). These data, together with the analysis of the ␣-domain portion required for recruitment of vinculin or F-actin on intracellular Shigella indicated that at least the N-terminal portion of VirG corresponding to amino acids 104 -506 was necessary for binding to vinculin, and that the vinculin binding ␣-domain portion coincided with the portion required for assembly of F-actin. DISCUSSION In this study, we have attempted to delineate the role of the different portions of VirG ␣-domain in the actin-based motility of S. flexneri, and reached the following conclusions: (i) the C-terminal ϳ220 amino acids of the ␣-domain are essential for unipolar deposition of VirG on the bacterial surface, (ii) the N-terminal two-thirds portion of the ␣-domain is required for assembly of F-actin on the bacterial surface in infected cells, (iii) VirG deposition on bacteria in infected cells elicits accumulation of vinculin at the same site, (iv) vinculin can interact with the two-thirds portion of the N-terminal ␣-domain in vivo and in vitro conditions, and (v) the vinculin head part is involved in the binding to VirG.
One of the VirG properties required for eliciting actin-based motility of Shigella in infected cells is an asymmetric distribution on the bacterial surface (8). Indeed, recent studies have shown that the rfa or galU mutant of S. flexneri, displaying a deep rough surface, has an aberrant distribution of VirG, unlike the asymmetric distribution on the wild-type bacterium (16,17). Consequently, those bacterial strains displayed the protein in a circumference tail fashion, rendering them defective in intercellular spreading. Similarly, when VirG was expressed at a very high level in a S. flexneri strain using a ptac-virG plasmid (14), the bacteria displayed aberrant surface distribution of VirG on the bacterial body, and became covered with F-actin around the surface in infected cells and defective Pellets were washed and subjected to SDS-PAGE and immunoblotting using anti-95 kDa head-specific monoclonal antibody VIN 11-5 or 30-kDa tail-specific monoclonal antibody 4-21 (upper panels) and rabbit anti-GST antibody (lower panels). In the case of the 30-kDa tail binding assay, the supernatant of the reaction mixture with GST-␣1 was also subjected to immunoblotting as a control. The reactivity of monoclonal antibody VIN 11-5 against purified 95-kDa head on the blotted nitrocellulose was same as the intact vinculin (data not shown). Bands smaller than the 95-kDa head were copurified with head protein, and are probably further digested products of V8 protease (28).
in the intercellular spreading. 2 Therefore, it is likely that a high level of surface-exposed VirG expression on Shigella results in a less defined asymmetry of VirG on the bacterial surface (20). In this study, we confirmed that the asymmetry of VirG distribution is an essential attribute for actin-based motility of Shigella in infected cells, since M94 (S. flexneri virG::Tn5) carrying pD10-1virG3 (the VirG variant lacking amino acids 509 -729), deposited VirG3 on the whole bacterial surface, leading to it becoming covered with F-actin and consequently defective in intercellular spreading. In contrast, the same bacteria expressing the VirG1 or VirG2 variant proteins possessing the C-terminal ␣-domain portion (see Fig. 1) showed asymmetric distribution like wild-type VirG, but no longer elicited F-actin assembly. To eliminate the possibility of the VirG ␤-core to contributing to the unipolar distribution of VirG, we investigated bacteria expressing a PhoA-␤-core fusion protein, in which the ␣-domain was replaced with PhoA, an E. coli alkaline phosphatase (12), for distribution of the fusion protein on the bacterial surface. Immunostaining with an anti-PhoA antibody revealed that the whole bacterial body was covered with PhoA. 2 We therefore conclude that VirG ␤-core, the outer membrane embedded VirG portion, is not directly involved in VirG distribution on the bacterial surface, but rather, that it is the preceding ϳ220 amino acids in the ␣-domain which contribute to asymmetric VirG distribution.
To date several proteins such as vinculin, plastin (fimbrin), and filamin have been reported to be associated with the Factin tail generated by intracellular S. flexneri (7,9), although the precise role of each protein in this process still remains to be elucidated at the molecular level. Vinculin, a protein linked to focal adhesions, has been shown to associate with the whole bacterial body within protrusions and the full-length of the F-actin tail as far as the protrusions (9). Plastin, an actin cross-linking protein, has been shown to be incorporated into the actin tail behind the bacterial body, thus contributing to the formation of a tight actin bundle (7). Filamin, an actin gelation protein, was shown to be weakly associated with the entire length of the F-actin tail (7). Among these proteins, vinculin was of particular interest to us, since, in addition to the above report (9), vinculin locates at focal adhesions and adhesion junctions, where it is thought to mediate attachment of actin filaments to integrin via talin or ␣-actinin (21,33). Recent studies indicated that vinculin consists of two distinct domains, the N-terminal 95-kDa head and C-terminal 30-kDa tail (28,32), and that vinculin binding to actin is mediated by the tail part, while the binding to talin is mediated by the head part (28,32). We thus wondered if the accumulation of F-actin and vinculin elicited around the VirG clot at one pole of the bacterium in infected cells could be a consequence of direct interaction of vinculin with VirG, analogous to vinculin binding to talin (28). To pursue the possible interaction between VirG and vinculin, we performed time course studies on VirG-vinculin-actin triple-labeled cells infected with bacteria expressing wild-type VirG or VirG3 variant. The recruitment of vinculin on the bacterial surface was confined to the site of VirG expression, and reorganization of vinculin on the bacterium appeared prior to the assembly of F-actin as seen in infected cells after 20 to 80 min of infection (Fig. 3). That a similar phenomenon occurred with intracellular S. flexneri was also stated by Kadurugamuwa et al. (9). These results were consistent with the observation by Geiger et al. (34), who suggested that focal adhesions and the associated vinculin serve as organizing centers for the assembly of actin-containing microfilament bundles.
The potential ability of vinculin to bind to VirG was investigated by two different approaches: the first was checking if vinculin in X. laevis egg extracts bound to VirG, and the second was the demonstration of direct binding of chicken gizzard vinculin to VirG. Since the Xenopus egg extracts have previously been shown to support actin-based motility of E. coli expressing VirG (18,19), we attempted to identify vinculin as a protein bound to VirG, and found that the GST-␣1 and GST-␣3 fusion proteins were bound by a 120-kDa protein which was subsequently identified as vinculin by immunoblotting with anti-chicken vinculin antibody (Fig. 5). Furthermore, chicken gizzard vinculin was shown to directly bind to the GST-␣1 and GST-␣3 fusion proteins. These data are thus comparable to the results that the two-thirds portion of the N-terminal ␣-domain is required for vinculin accumulation on Shigella in infected cells.
The two-thirds N-terminal portion contains at least eight repetitive sequences rich in Gly (see Fig. 1) (12). According to our preliminary data, the Gly-rich repeats seem to be a component of VirG involved in eliciting F-actin assembly on Shigella in infected cells. Bacteria expressing a VirG variant lacking the first Gly-rich repeat proximal to the N terminus of the ␣-domain (corresponding to VirG amino acids 104 to 158), resulted in a significantly shorter F-actin tail than that produced by the wild-type VirG. Furthermore, bacteria expressing another VirG variant lacking the first two Gly-rich repeats (corresponding to VirG amino acids 104 to 226), completely failed to elicit F-actin assembly. 2 Interestingly, the ActA protein of Listeria monocytogenes and the IactA protein of Listeria ivanovii, required for actin-based motility in infected cells, possess four and seven Pro-rich repeats in the middle of each polypeptide, respectively (35,36). Those repeats are essential for the bacteria to elicit full F-actin assembly in mammalian cells (37), since bacteria expressing an ActA variant lacking the Pro-rich repeats lost their ability to elongate the F-actin tail (37). Chakraborty et al. (39) found that the Pro-rich repeat domains are bound by vasodilator-stimulated phosphoprotein, a novel protein associated with focal adhesions and able to interact with microfilaments (38), and showed that the binding is required for the elongation of F-actin at one pole of the bacterium in infected cells (39), although the precise role of vasodilator-stimulated phosphoprotein in eliciting F-actin assembly on Listeria in infected cells still remains to be ascertained. In their study, they noted that vasodilator-stimulated phosphoprotein was able to associate with actin tail from intracellular S. flexneri, but were unable to detect direct binding to VirG in an in vitro study. Therefore, it is less likely that the virG ␣-domain portion containing Gly-rich repeats plays a role in recruitment of F-actin via vasodilator-stimulated phosphoprotein, but rather, contributes to recruitment of F-actin through interaction with another protein, such as vinculin, able to directly bind to F-actin.
The above scenario may be possible, since we have observed FIG. 7. A possible model for recruitment of F-actin assembly on the surface-exposed VirG moiety on Shigella in infected cells. VirG ␣-domain exposed on the bacterial surface recruits vinculin, and this may lead to unmasking of the intramolecular vinculin headtail interaction (32). The vinculin tail portion could then interact with F-actin or/and other signaling molecules (21,44). Multimerization of vinculin may occur through self-association or localized accumulation of VirG, which may promote cross-linking actin filaments (31,32). in this study that the 95-kDa vinculin head part binds specifically to VirG. Indeed, when the vinculin head and the tail parts were obtained from intact vinculin by digestion with V8 protease, only the head part bound to the GST-␣1 or GST-␣3 (Fig. 6B). In that experiment, we noted differing affinities of the intact vinculin molecule and the vinculin head part to VirG, since when equivalent moles of intact vinculin and the head part were incubated with GST-␣1 or GST-␣3, the amount of bound intact vinculin was significantly lower compared to the bound vinculin head as determined by immunoblots (see Fig. 6,  A and B). These differences were reproducible among experiments. This was thus reminiscent of the recent studies by Johnson and Craig (28,32) who demonstrated that an intramolecular interaction between the head and tail part of vinculin masks the site in the purified protein responsible for the binding to talin and actin. Although we must be careful in drawing our conclusions, it is tempting to speculate that the observed differing affinities between the intact vinculin and the head part to VirG could result from the separation of the vinculin head from the intact vinculin with V8 protease. As suggested by previous study, the intramolecular interaction of vinculin can be prevented through the interaction between vinculin and talin (28). In this regard, although there is no significant similarity between the amino acids of the ␣-domain and talin (40), a conformational change in vinculin evoked upon interaction with VirG may promote the interaction of vinculin with F-actin near the bacterial surface in infected cells. Alternatively, an intracellular signal evoked by the infection of Shigella may cause a conformational change in the vinculin molecule such as by binding with tyrosine-phosphorylated paxillin or phosphatidylinositol-4,5-bisphosphate (41)(42)(43), which may lead to the binding of the vinculin tail to F-actin (44).
A previous study revealed that a 95-kDa processed form of VirG released from S. flexneri, corresponding to the ␣-domain, can interact with F-actin along the actin tail in the infected cells (8). However, this was recently shown to be the consequence of cross-reactivity of the VirG monoclonal antibody with another as yet uncharacterized 70-kDa host protein associated with the F-actin tail (15). Since there has been no conclusive evidence supporting the ability of VirG to directly interact with F-actin (20), it is thus not probable that VirG can directly recruit F-actin assembly on the bacterial surface in infected cells, but rather, that the recruitment of F-actin by VirG is mediated by an actin-binding protein such as vinculin as proposed in Fig. 7. In any case, further studies must be awaited to discover the precise role of vinculin-VirG interaction in actinbased motility of Shigella in infected mammalian cells.