The Enterohemorrhagic Escherichia coli Effector EspW Triggers Actin Remodeling in a Rac1-Dependent Manner

ABSTRACT Enterohemorrhagic Escherichia coli (EHEC) is a diarrheagenic pathogen that colonizes the gut mucosa and induces attaching-and-effacing lesions. EHEC employs a type III secretion system (T3SS) to translocate 50 effector proteins that hijack and manipulate host cell signaling pathways, which allow bacterial colonization and subversion of immune responses and disease progression. The aim of this study was to characterize the T3SS effector EspW. We found espW in the sequenced O157:H7 and non-O157 EHEC strains as well as in Shigella boydii. Furthermore, a truncated version of EspW, containing the first 206 residues, is present in EPEC strains belonging to serotype O55:H7. Screening a collection of clinical EPEC isolates revealed that espW is present in 52% of the tested strains. We report that EspW modulates actin dynamics in a Rac1-dependent manner. Ectopic expression of EspW results in formation of unique membrane protrusions. Infection of Swiss cells with an EHEC espW deletion mutant induces a cell shrinkage phenotype that could be rescued by Rac1 activation via expression of the bacterial guanine nucleotide exchange factor, EspT. Furthermore, using a yeast two-hybrid screen, we identified the motor protein Kif15 as a potential interacting partner of EspW. Kif15 and EspW colocalized in cotransfected cells, while ectopically expressed Kif15 localized to the actin pedestals following EHEC infection. The data suggest that Kif15 recruits EspW to the site of bacterial attachment, which in turn activates Rac1, resulting in modifications of the actin cytoskeleton that are essential to maintain cell shape during infection.

The actin cytoskeleton, which is targeted by many bacterial pathogens, is essential for cell integrity, motility, membrane trafficking, and shape changes (14). Rho GTPases, which belong to the family of Ras-related small GTPases, are key regulators of various cellular processes, including actin polymerization, microtubule dynamics, vesicle trafficking, cell polarity, and cytokinesis (15). The best-characterized members of the Rho GTPase family are RhoA, Rac1, and Cdc42, the activation of which leads to the assembly of stress fibers, lamellipodia/ruffles, and filopodia, respectively (16). Switching of Rho GTPases from an inactive GDP-bound state to an active GTP-bound state is mediated by guanine nucleotide exchange factors (GEFs). The switch back from the active GTP to an inactive GDP-bound state is regulated by GTPase-activating proteins (GAPs). In their GTP-bound conformation, Rho GTPases interact with and activate downstream target effectors, such as serine/threonine kinases, tyrosine kinases, lipid kinases, lipases, oxidases, and scaffold proteins (17). As Rho GTPases are important regulators of the actin cytoskeleton, bacterial pathogens have evolved strategies to subvert their signaling during infection.
Bacterial guanine nucleotide exchange factors, which belong to the SopE family, act as bacterial Rho GEFs to activate the host Rho GTPase (18). The A/E pathogen effector Map induces filopodia via Cdc42 at the site of attachment (19,20), EspM promotes stress fibers via RhoA activation (21), and EspT triggers ruffle and lamellipodia formation by Rac1 (22). A/E pathogens also translocate effectors that inactivate Rho GTPases. EspH globally inactivates DH-PH domain mammalian Rho-GEFs but not the bacterial Rho-GEFs (23). Tir antagonizes the activity of Map as it downregulates formation of filopodia (24), while EspO2 interacts with EspM2 and blocks formation of the stress fibers (25).
Using a transfection-based screen, we recently identified EspW EHEC as a regulator of actin filament organization. EspW has been shown previously to be secreted by EHEC and translocated into mammalian cells in a type 3-dependent manner (26). However, until now, no function has been identified for this effector. The aim of this study was to investigate the role of EspW during EHEC infection and its putative role as a Rho GTPase regulator.

RESULTS
Screening of espW in EPEC clinical isolates. EspW is a 352-amino-acid effector and is located in the SP17 pathogenic island, which also encodes EspM2 and members of the NleG family (see Fig. S1A in the supplemental material). So far, EspW has been reported only in EHEC O157:H7 and EPEC B171 (O111:H Ϫ ) strains, with no homologs among other bacterial species. Using the BLAST algorithm with EspW as the index protein, we confirmed that it was present in the sequenced EHEC O157:H7 strains, in five non-O157:H7 EHEC strains (O111:H Ϫ , O111:H11, O26:H11, O103:H2, and O103:H25), and in Shigella boydii (Fig. S1B). Furthermore, a putative coding sequence for a truncated version of EspW containing the N-terminal 206 amino acids (EspW 1-206 ) was present in two EPEC strains (CB9615 and RM12579) belonging to serotype O55:H7 (Fig.  S2), a progenitor of EHEC O157:H7 (27). In order to determine if either the long or short versions of espW are present in other EPEC strains, we screened by PCR a collection of 132 clinical isolates available in our laboratory. This revealed that the long version of espW is present in 52% of the tested stains (Table 1). Furthermore, espW  was found in 10 of the 132 (8%) strains tested (Table 1). Interestingly, 9 of the 10 espW 1-206 genes belonged to serotype O55:H7. Neither of the espW variants was found in C. rodentium and the prototype EPEC strain E2348/69, while the prototype atypical EPEC strain E110019 (O111:H9) contains the long version of espW.
EspW interacts with the C terminus of Kif15. In order to identify the EspW host cell partner protein, we performed a yeast two-hybrid screen using a HeLa cell cDNA library as bait and identified the carboxy terminus of Kif15, Kif15 1092-1368 , as a putative partner. The interaction was confirmed by direct yeast two-hybrid (DY2H). Importantly, Kif15 1092-1368 interacted with the full-length EspW (Fig. 1B) but not with EspW  . To further map the binding site of EspW to Kif15, five Kif15 truncation fragments were generated and tested by DY2H (Fig. 1A). An empty pGAD-T7 plasmid was used as a negative control. No growth was observed on selected media (QDO) when yeast were cotransformed with EspW and Kif15 1142-1347 , Kif15 1142-1368 , or the negative control. In contrast, growth was seen following cotransformation with EspW and Kif15 1092-1347 or Kif15 1092-1142 (Fig. 1C), suggesting that the C-terminus coil-coil domain of Kif15 plays an important role in the interaction with EspW.
Kif15 localizes to the pedestals and colocalizes with EspW. We aimed to determine the localization of Kif15 during EHEC infection. However, we were unable to detect endogenous Kif15, and localization of overexpressed Kif15 was difficult to detect due to poor transfection efficiency. Accordingly, we determined the localization of ectopically expressed Kif15 1092-1368 , used in the DY2H, following EHEC infection of transfected Swiss 3T3 cells. Cells expressing myc-green fluorescent protein (GFP) were used as a negative control. Immunofluorescence (IF) microscopy revealed that Kif15 1092-1368 , but not GFP, localized to the actin pedestals at the site of EHEC attachment (Fig. 1D). Interestingly, cells transfected with Kif15 1092-1368 and infected with an EHEC ΔespW strain present a similar recruitment of Kif15 1092-1368 into the pedestal (Fig. S3), suggesting EspW is not required for localization of Kif15 to the pedestal.
Deletion of espW induces cell shrinkage that could be overcome by Rac1 activation. To assess the role of EspW during infection, cells were infected for 3 h with wild-type (WT) EHEC, an EHEC ΔespW strain, or an EHEC ΔespW strain complemented with pEspW. Immunofluorescence reveals that infection with the EHEC ΔespW strain induced significant cell shrinkage (56%) compared to infection with WT EHEC (12%) (Fig. 4A). Partial complementation was observed for the cell infected with the EHEC ΔespW strain complemented with pEspW (32%) (Fig. 4C).
To determine if the cell shrinkage was linked with lack of activation of Rac1, cells were infected with an EHEC ΔespW strain overexpressing EspT, an effector known to activate Rac1 (22). The EspT W/A mutant, lacking the GEF activity of EspT, was used as a negative control (Fig. 4B). Expression of WT EspT significantly reduced cell shrinkage (31%) compared with cells infected with the EHEC ΔespW strain complemented with pEspT W/A (50%) (Fig. 4C).
In order to confirm that the cell shrinkage was caused by the lack of Rac1 activation, we chemically induced activation of Rac1 during infection by adding 100 nM sphin- gosine 1-phosphate (S1P) to the culture medium (28) and quantified the number of shrunken cells after infection ( Fig. 5A and B). S1P treatment significantly reduced shrinking of cells infected with the EHEC ΔespW strain from 53% to 33% (Fig. 5B). These results suggest that EspW activates Rac-1, which stabilizes the shape of infected cells.

DISCUSSION
In this study, we found that espW is common among clinical EHEC and EPEC isolates; an espW orthologue is also found in Shigella boydii. The majority of the EPEC strains contain the full-length espW gene, while others, mainly belonging to EPEC O55:H7, encode a truncated EspW isoform. Although the truncated form of EspW does not induce actin reorganization, it is possible that it has other biological functions.
Using a two-hybrid screen, we identified Kif15 as a specific partner of the full-length EspW isoform. Human Kif15 is a multimeric protein of 1,388 amino acids which belongs to the kinesin family (29). It has an N-terminal motor domain (residues 19 to 375) followed by a long alpha-helical rod-shaped stalk predicted to form an interrupted coiled coil. The C-terminal region has been shown to contain a putative actin interacting region (residues 743 to 1333) (30). Moreover, in HeLa cells, Kif15 has been shown to concentrate on spindle poles and microtubules in early mitosis and to localize with actin in late mitosis (31). One possibility is that Kif15 switches binding from one filament system to the other, while another possibility is that Kif15 associates with the most abundant cytoskeletal filament system (31). In this study, we mapped the EspW binding site to a segment of Kif15, amino acids 1092 to 1142. This segment is a known binding site for both Ki-67 (1017 to 1237) and actin (743 to 1333). The exact role of Kif15 during infection is still unclear, as labeling of EspW in EPEC did not allow us to localize the effector during infection. However, its recruitment to the pedestal during EPEC infection is independent of EspW. We therefore hypothesize that Kif15 recruits EspW and determines its spatial distribution, similar to the function of NHERF1 or NHERF2 toward the effector Map (32).
EPEC and EHEC, like many other enteric pathogens, target actin cytoskeleton as part of their infection strategy. The hallmark of EPEC and EHEC infection of cultured cells is formation of actin pedestal-like structures underneath the attached bacteria. In EPEC, formation of these structures is dependent on the effector Tir and activation of N-WASP and independent of activation of mammalian Rho GTPases (33). However, EspH, which is a global inhibitor of endogenous mammalian GEFs (23), is required for efficient actin pedestal elongation (34), suggesting that Rho GTPases are partially involved in this process. Importantly, EPEC and EHEC translocate several effectors, belonging to the SopE family, which have a GEF activity toward mammalian Rho GTPases (18). In vitro, EspT, which activates Rac1, triggers formation of ruffles or lamelipodia, and in vivo it induces expression of KC and tumor necrosis factor alpha (TNF-␣) (35). In this study, we found that EspW also appears to activate Rac1, either directly or indirectly, in a compartmentalized fashion; this is in contrast to EspT, which has a more global effect. Nonetheless, the phenotype of espW deletion could be partially complemented by espT, suggesting some activity overlap. Due to poor solubility, we were not able to identify whether EspW directly activates Rac1. Importantly, multiple biological systems revealed that activation or inhibition of the Rho GTPase has to be fine-tuned both spatially and temporally. Their overactivation or inhibition have detrimental effects leading to activation of alarm signals (36) or apoptosis (37). During EPEC infection, activation of Cdc42 is limited to the bacterial binding sites (19), followed by rapid inhibition by Tir (19). The effector EspO, expressed by a selection of EPEC and EHEC strains, has been reported to inactivate EspM2 (RhoA GEF). Interestingly, deletion of espO1 and espO2 leads to cell shrinkage in an EspM2-dependent manner (25). Rac1 and RhoA have antagonistic effects (38). Interestingly, we found that cells infected with EHEC expressing EspM1 and EspM2 but missing EspW undergo cell shrinkage. This cell shrinkage phenotype was not associated with decreased cell attachment or with any signs of cell death, including nucleus condensation, loss of membrane permeability, or membrane blebbing, for the duration of the experiment. Interestingly, we found that EPEC and EHEC strains expressing EspM also express either EspT or EspW, suggesting that activation of RhoA and Rac1 need to be coordinated during infection. Furthermore, deletion of Rac1 impairs focal adhesion complex formation and cell spreading (39). Taken together, these observations suggest that EPEC and EHEC have developed a complex mechanism to control cell shape by manipulating the localization and activation of RhoA and Rac1. Any dysregulation leading to an uncontrolled activation leads to dramatic cell morphology changes. Further studies will be needed in order to understand the spatiotemporal regulation of the Rho GTPase during EPEC and EHEC infections.

MATERIALS AND METHODS
Bacterial strains, growth conditions, and cell culture. The bacterial strains used in this study and their origins are listed in Table 2. Bacteria were grown from a single colony in Luria-Bertani (LB) broth in a shaking incubator (200 rpm) at 37°C for 18 h or on agar supplemented with ampicillin (100 g/ml) or kanamycin (50 g/ml). For cell infections, EHEC strains were grown in LB in a shaking incubator (200 rpm) at 37°C for 8 h and then subcultured (1/500) in Dulbecco's modified Eagle's medium (DMEM) with low glucose and grown overnight at 37°C without agitation in a 5% CO 2 incubator (primed culture).
Saccharomyces cerevisiae (AH109) was grown in YPDA medium (20 g/liter Difco peptone, 10 g/liter yeast extract, 2% glucose, and 0.003% adenine hemisulfate) for 48 h at 30°C. For the yeast two-hybrid screen, clones containing interaction partners were selected on high-stringency quadruple-dropout (QDO) medium lacking leucine, tryptophan, histidine, and adenine in the presence of X-␣-Gal (Clontech Laboratories, Inc.). Successful transformation with bait and prey plasmids was selected by plating on double-dropout (DDO) medium lacking leucine and tryptophan. Bait-prey interactions were assessed by streaking the transformed clones from DDO onto QDO selection medium.
Plasmids and molecular techniques. Plasmids used in this study are listed in Table 2, and primers are listed in Table S1 in the supplemental material.
The EHEC ΔespW (ICC1111) strain was generated using a lambda red-based mutagenesis system (40) in which espW was replaced by a kanamycin cassette. Plasmid pSB315 was the source of the kanamycin resistance gene (aphT), which was purified following EcoRI restriction digestion. Primer pair P23/P24 was used to PCR amplify espW with 500-bp upstream and downstream flanking regions from E. coli O157:H7 (85-170) genomic DNA. The PCR product was cloned into TOPO Blunt II vector (Invitrogen), and espW was removed by inverse PCR using the primer pair P25/P26. The linear PCR product was then EcoRI digested to allow ligation of the kanamycin cassette. The insert was then amplified using the primer pair P23/P24 and the PCR product electroporated into WT EHEC containing pKD46 encoding the lambda red recombinase. Transformants were selected on kanamycin plates, and the deletion of espW was confirmed by PCR and DNA sequencing (using primer pair P27/P28).
Yeast two-hybrid assays. Yeast two-hybrid screening using EspW as prey and a cDNA library as bait was performed as described previously (43). Briefly, a pretransformed MATCHMAKER HeLa cell cDNA library (Clontech) was screened according to the manufacturer's protocol for proteins interacting with EspW. The lithium acetate method was used to transform pGBT9-espW (pICC1714) ( Table 2) into yeast strain AH109 (MATa), and transformants were selected on Trp-minus-synthetic-defined agar plates. Following mating with the Y187 (MAT␣) yeast strain containing the cDNA library, diploids cells were selected on DDO and QDO media for selection of protein interactions. The cDNA-containing pGADT7 plasmid was rescued from positive clones and the cDNA identified by DNA sequencing. The prey plasmid and derivatives (Table 2) were then retransformed into AH109 either on its own to determine possible self-activation or with pICC1714 or pICC1715 to confirm interaction by direct yeast two-hybrid assay. Infection of Swiss 3T3 and HeLa cells. Forty-eight hours prior to infection, Swiss 3T3 or HeLa cells were seeded in 24-well plates containing 13-mm glass coverslips (VWR International) at a density of 5 ϫ 10 5 cells per well. Before infection, the cells were washed 3 times with phosphate-buffered saline (PBS) and the medium was replaced with fresh DMEM without FCS. Cells in 24-well plates were infected with 20 l of primed cultures. The plates were then centrifuged at 200 rpm for 5 min at room temperature, and infections were carried out for 3 h at 37°C in 5% CO 2 without agitation. After infection, monolayers were washed at least 10 times in PBS to remove the bacteria and were fixed for immunofluorescence (to assess cell morphology) as described below.
SEM. For scanning electron microscopy (SEM), cells were washed 3 times in phosphate buffer, pH 7.4, and then fixed with 2.5% glutaraldehyde in phosphate buffer, pH 7.4. Cells were washed 3 times in phosphate buffer before being postfixed in 1% osmium tetroxide for 1 h. Cells were then washed 3 times in phosphate buffer and dehydrated for 15 min in graded ethanol solutions from 50% to 100%. The cells were then transferred to an Emitech K850 critical point drier and processed according to the manufacturer's instructions. The coverslips were coated in gold/palladium mix using an Emitech Sc762 mini sputter. Samples for SEM were then examined blindly at an accelerating voltage of 25 kV using a JEOL JSM-6390.
Statistical analysis. All data were analyzed with GraphPad Prism software, using one-way analysis of variance (ANOVA). Results were expressed as means and standard deviations (SD). Statistical significance was determined by a two-tailed Student t test. A P value of Ͻ0.05 was considered significant.

ACKNOWLEDGMENTS
This work was supported by grants from The Wellcome Trust, the BBSRC, and MRC.