Modulation of Host Cytoskeleton Function by the Enteropathogenic Escherichia coli and Citrobacter rodentium Effector Protein EspG

Abstract

EspG is a conserved protein encoded by the locus of enterocyte effacement (LEE) of attaching and effacing (A/E) pathogens, including enteropathogenic and enterohemorrhagic Escherichia coli and Citrobacter rodentium. EspG is delivered into infected host cells by a type III secretion system. The role of EspG in virulence has not yet been defined. Here we describe experiments that probe the virulence characteristics and biological activities of EspG in vitro and in vivo. A C. rodentium espG mutant displayed a significantly reduced ability to colonize C57BL/6 mice and to cause colonic hyperplasia. Epitope-tagged EspG was detected in the apical regions of infected colonic epithelial cells in infected mice, partially localizing with another LEE-encoded effector protein, Tir. EspG was found to interact with mammalian tubulin in both genetic screens and gel overlay assays. Binding to tubulin by EspG caused localized microtubule depolymerization, resulting in actin stress fiber formation through an undefined mechanism. Heterologous expression of EspG in yeast resulted in loss of cytoplasmic microtubule structure and function, preventing coordination between bud development and nuclear division. Yeast expressing EspG were also unable to control cortical actin polarity. We suggest that EspG contributes to the ability of A/E pathogens to establish infection through a modulation of the host cytoskeleton involving transient microtubule destruction and actin polymerization in a manner akin to the Shigella flexneri VirA protein.

Pathogenic strains of Escherichia coli are responsible for many diseases, including meningitis, urinary tract infections, and diarrhea (4). Enteropathogenic E. coli (EPEC) is a leading cause of bacterium-mediated diarrhea in children and a major endemic health threat in the developing world and is responsible each year for the death of several hundred thousand people. Isolated outbreaks also occur in children in developed countries. EPEC binds to intestinal epithelial cells, forming an attaching and effacing (A/E) lesion resulting from localized microvillus destruction and the formation of a pedestal-like projection composed of epithelium-derived cytoskeletal components (9).

Many of the characterized bacterial factors responsible for the formation of A/E lesions and resultant disease are both encoded and regulated by a pathogenicity island described as the locus of enterocyte effacement (LEE). The LEE encodes a type III secretion system, which serves as a molecular syringe to direct the active transport of proteins from the bacterial cytoplasm into the cytoplasm of an associated eukaryotic cell (15). The proteins injected into the host cell are called effectors. The bacterial protein Tir is inserted into the host cell plasma membrane to serve as the receptor for the intimin protein (28). EspF has been shown to disrupt intestinal barrier function (22), and Map is translocated to host mitochondria (19). EspH may play a role in filopodia formation through interference with the actin cytoskeleton (30). Recently, effectors encoded outside of the LEE have been described (6), including NleA, which localizes to the host Golgi apparatus (13).

An additional effector, EspG, has been identified, but its function has yet to be elucidated. Elliott et al. described the protein as translocated to human epithelial cells during infection (11). An EPEC ΔespG mutant had no defect in in vitro assays of virulence, but a rabbit EPEC ΔespG strain displayed diminished intestinal colonization. It was noted that EspG exhibits 40% amino acid sequence similarity to the Shigella flexneri VirA protein, which is involved in destabilizing microtubules to promote membrane ruffling (34). Intracellular persistence of a Shigella ΔvirA mutant could be restored if complemented with espG, suggesting functional conservation (11). Thus, EspG may have an accessory role in colonization via modulation of the host cytoskeleton. It is interesting that the two proteins may have similar biochemical function, given the intracellular lifestyle of Shigella versus the predominantly extracellular pathogenesis of EPEC.

There is strong homology among the LEE genes in pathogenic Escherichia coli and the mouse A/E pathogen Citrobacter rodentium (5). The use of C. rodentium as a small-animal model for studying human A/E pathogens has facilitated study of several virulence factors in clinically relevant pathogens. Mice infected with C. rodentium develop mild diarrhea, A/E lesions, pedestal-like projections, and colonic hyperplasia, making this system an excellent model for EPEC and EHEC infection of the human gut.

The present study utilizes the C. rodentium model of EPEC infection to assess the virulence of a ΔespG strain and the in vivo localization of EspG in the host epithelium. We also assess the function of EspG through two-hybrid library screening, immunofluorescence, and the heterologous expression of EspG in Saccharomyces cerevisiae. The results suggest a role for EspG in early host colonization, potentially through modulation of epithelial microtubule and actin structures.

MATERIALS AND METHODS

Generation of a nonpolar C. rodentium espG deletion mutant.

A PCR fragment containing Citrobacter rorf1 and espG/rorf2 genes, as well as a 1.1-kb flanking region downstream of the rorf1 gene and a 0.5-kb flanking region upstream of the espG gene, was cloned into pCRII-TOPO to create pCRII-espG/rorf1. The primers used for the PCR were 5′-TGAGTGGAGTTACAACGTC and 5′-CGGCGACGGAATAACCGCGTTCCG. The plasmid pCRII-espG/rorf1 was used as a template for inverse PCR using Elongase with primers 5′-GCCGTCGACGTATGGCATGTGGCGATTGATGGG and 5′-CCCGTCGACGCCTCATCTGAAACATATCTGAAC to create an internal deletion (codons 152 to 388) in espG. A SalI site was introduced into the site of deletion. The inverse PCR product was digested with SalI, and gel purified before self-ligation and transformation into DH10B. The espG gene with the internal deletion was subcloned as a KpnI/XbaI fragment into the suicide vector pRE118, along with its flanking regions, and used for allelic exchange in wt C. rodentium as described previously (7). The Citrobacter espG deletion mutants were verified by PCR. (See Table 1 for a description of strains and plasmids discussed in this study.)

TABLE 1.

Strains and plasmids used in this study

Strain or plasmid	Genotype or descriptiona	Source or reference
Strains
E. coli
EPEC E2348/69	wt O127:H6 isolate	20
sPRH232	E2348/69/pPRH282	This study
EPEC JPN15	E2348/69 cured of pMAR2 plasmid	1, 17
sPRH2	EPEC ΔespG; Nalr	This study
sPRH247	EPEC ΔespG/pPRH282	This study
BL21 (DE3)	F−ompT hsdSB (rB− mB−) gal dcm (DE3)	Novagen
sPRH22	BL21(DE3)/pPRH10	This study
C. rodentium
DBS100	ATCC 51459	29
sPRH233	DBS100/pPRH282	This study
sPRH133	DBS100 ΔescF	This study
sPRH234	DBS100 ΔescF/pPRH282	This study
S. cerevisiae         YPH499	MATaade 2-101 trp1-63 leu2-1 ura3-52 his3-200 lys2-801	P. Hieter
Plasmids
pBT	Bacterial two-hybrid bait vector	Stratagene
pTRG	Bacterial two-hybrid target vector	Stratagene
pTuba6	Human alpha-tubulin clone	Orbigen
pPRH18	EPEC espG in pBT	This study
pTOPO-2HA	pCR2.1-TOPO-based expression vector for C-terminal 2HA2 tagging; Ampr Kanr	6
pPRH84	tuba6 in pTRG	This study
pCVD442	suicide plasmid for allelic exchange	8
pCR2.1-TOPO	PCR cloning vector; Ampr Kanr	Invitrogen
pPRH295	PCR2.1.TOPO-espG	This study
pPRH97	PCRII-TOPO-Citro espG/rorf1	This study
pRE118	sacB-based suicide vector for allelic exchange; Kanr	10
pPRH294	pRE118-CitroΔespG KpnI/Xbal	This study
pET-15b	His-tag protein expression vector Ampr	Novagen
pPRH10	EPEC espG in pET-15b	This study
pPRH282	Citrobacter espG-HA2	This study
pPRH1	pCMV-Tag2B, FLAG expression vector; Kanr	Stratagene
pPRH7	EPEC espG in pCMV-Tag2B	This study
pPRH228	EPEC espG(V220-Q332) in pCMV-Tag2B	This study
pEG(KG)	Galactose-inducible vector for conditional expression of GST fusions	25
pPRH253	pEG(KG)-espG	This study

^aNal^r, nalidixic acid resistant; Amp^r, ampicillin resistant; Kan^r, kanamycin resistant.

Generation of a nonpolar EPEC espG deletion mutant.

A PCR fragment (1.3-kb) containing the espG gene with short flanking sequences on both sides of espG was cloned into pCR2.1-TOPO to create pTOPO-espG. The primers used for the PCR were 5′-TTGTGGGTGTGATTTCTATTATTGG and 5′-CTTTCTATGCGTTTAAAACACCCGC. The PCR fragment in pTOPO-espG was subcloned into pBluescript II SK(+) as an EcoRI fragment to create pBluescript-espG, and this plasmid was used for inverse PCR with deletion primers 5′-GAAGATCTGCGCGATGCACTAGGACTTAAA and 5′-GTAGATCTGGCACGATGAACCTGAATGTA to create an internal deletion (codons 159 to 230) in espG. A BglII site was introduced into the site of deletion. The inverse PCR product was digested with BglII, and gel purified before being self-ligated and transformed into DH10B. The espG gene with the internal deletion was subcloned as a SalI/SacI fragment into the suicide vector pCVD442 (SalI/SacI), along with its flanking regions, and used for allelic exchange in an EPEC nalidixic acid (Nal)-resistant derivative. Briefly, pCVD-ΔespG was purified from Escherichia coli SY327λpir, transformed into E. coli SM10λpir by electroporation, and then transferred into the EPEC Nal-resistant derivative by conjugation. Sucrose selection was then carried out to select for EPEC espG deletion mutants as described previously (8). Mutants were verified by PCR.

Mouse infections.

Four-week-old C57BL/6 or C3H/HeJ mice from Jackson Laboratory were infected by oral gavage with 100 μl (∼5 × 10⁸ CFU) of the indicated C. rodentium strain and analyzed at day 6 and day 10 postinfection (p.i.) as described previously (7). The data were analyzed with the Student t test and deemed significant if P was <0.05.

Adherence and invasion assays in HeLa cells.

A total of 10⁵ HeLa cells/well were seeded in a 24-well tissue culture plate and grown in Dulbecco modified Eagle medium overnight. For adherence assays, HeLa cells were infected for 1.5 h with 5 μl of an EPEC standing culture grown overnight in 2 ml of LB for a multiplicity of infection of ∼100:1. The cells were washed 10 times with phosphate-buffered saline (PBS) and lysed in 1% Triton X-100 (vol/vol) in water for 5 min. The samples were serially diluted and plated on LB plates to determine the number of adherent bacteria. For invasion assays, HeLa cells were infected similarly for 3 h, washed three times with PBS, and incubated in DMEM containing 10% fetal calf serum and 100 μg/ml of gentamicin for 1 h to kill extracellular bacteria. The cells were washed three times with PBS, lysed in 1% Triton X-100, and plated as described above. Assays were performed twice in triplicate.

In vivo immunofluorescence staining.

C3H/HeJ mice were orally inoculated with C. rodentium ΔespG carrying a plasmid expressing espG-HA₂ or with wild-type (wt) C. rodentium. For inoculations, bacteria were grown overnight in Luria broth, with cultures of the C. rodentium ΔespG carrying the appropriate antibiotic. Mice were infected by oral gavage by using 0.1 ml of Luria broth containing ∼2.5 × 10⁸ CFU of C. rodentium. Four days later the mice were euthanized, and colonic tissues were collected, rinsed with ice-cold PBS, embedded in optimal cutting template (OCT) compound (Sakura Finetech), frozen with isopentane (Sigma) and liquid N₂, and stored at −70°C. Serial sections were cut 6 μm thick and fixed in ice-cold acetone for 10 min. Tissues were blocked by using 1% bovine serum albumin. To identify infected epithelial cells, we used rat polyclonal antisera generated against C. rodentium translocated intimin receptor (Tir; 1:5,000), followed by an Alexa488-conjugated goat anti-rat as the secondary antibody (1:300; Molecular Probes). EspG-HA₂ was stained by using biotinylated monoclonal α-HA (clone 16B12, 1:300; Covance), followed by Alexa594-conjugated streptavidin (1:500; Molecular Probes). In addition, 1 μg of 4′,6′-diamidino-2-phenylindole (DAPI)/ml was used to stain host cell DNA. Coverslips were mounted in Mowiol (Aldrich) and viewed at 350, 488, and 594 nm on a Zeiss Axiophot epifluorescence microscope.

Bacterial two-hybrid assay.

EPEC espG was subcloned into the BacterioMatch two-hybrid system vector kit bait plasmid (Stratagene). A mouse spleen cDNA library (Clontech) was subcloned into the two-hybrid target plasmid. Library screening was conducted according to the manufacturer's instructions, with an optimized carbenicillin concentration of 250 μg/ml. A cDNA clone of human alpha tubulin was purchased from Orbigen. β-Galactosidase assays were performed as described previously (23).

In vitro immunofluorescence staining.

Caco-2 cells were grown on glass coverslips in 24-well tissue culture plates and infected for 4 h with 1.0 μl of bacterial overnight culture. After infection, cells were washed three times in PBS containing Ca²⁺ and Mg²⁺ and fixed in 2.5% paraformaldehyde in PBS for 10 min at room temperature. Cells were permeabilized in 0.1% saponin in PBS, blocked in 5% goat serum in PBS-0.1% saponin, and incubated with the following antibodies diluted 1:1,000 in blocking solution for 1 h at room temperature: mouse anti-α-tubulin antibody (Sigma), mouse anti-β-tubulin antibody (Sigma), and rat anti-HA antibody and an Alexa488-phalloidin stain (Molecular Probes). After addition of the primary antibody, the cells were washed extensively with PBS and probed with Alexa dye-conjugated antibodies, and DAPI (1 μg/ml; Sigma) was used to stain the bacterial and host cell DNA. The coverslips were mounted in Mowiol (Aldrich) and viewed at 350, 488, and 594 nm on a Zeiss Axiophot epifluorescence microscope. Transfection experiments were performed with 0.5 μg of plasmid suspended in Fugene as recommended by the manufacturer (Roche) and immunostained 36 h posttransfection.

His-EspG purification.

EPEC espG was subcloned into pET-15b and expressed in BL21(DE3). Next, 250 ml of bacterial culture was grown to an optical density at 600 nm of 0.6, and IPTG (isopropyl-β-d-thiogalactopyranoside) was added to 1 mM. After 3 h additional growth, cells were pelleted and resuspended in sonication buffer (5 mM imidazole, 6 M guanidine, 250 mM NaCl, 20 mM Tris-HCl [pH 7.9]). Cells were sonicated and centrifuged to clarify the supernatant. The supernatant was added to packed preequilibrated Ni-nitrilotriacetic acid (NTA) agarose and incubated for 1 h at 4°C. The slurry was poured into a disposable Poly-Prep chromatography column and washed five times with wash buffer (60 mM imidazole, 6 M guanidine, 250 mM NaCl, 20 mM Tris-HCl [pH 7.9]). His-EspG was eluted in 1-ml fractions of wash buffer supplemented with increasing concentrations of imidazole and analyzed by sodium dodecyl sulfate-10% polyacrylamide gel electrophoresis (SDS-10% PAGE). Fractions containing His-EspG were dialyzed into storage buffer (25 mM sodium phosphate [pH 7.5], 20 mM NaCl, 1 mM dithiothreitol, 5% glycerol) by using a Slide-A-Lyzer dialysis cassette (Pierce).

Gel overlays.

Purified His-EspG and tubulin (Cytoskeleton) were resolved by SDS-8% PAGE, transferred to nitrocellulose, and processed as previously described (12). Blots were probed with either anti-His (1:1,000) or anti-tubulin (1:1,000) primary antibody and goat anti-mouse horseradish peroxidase-conjugated secondary antibody (1:5,000), followed by detection with ECL reagent (Amersham).

Expression and phenotypic analysis in Saccharomyces cerevisiae.

The S. cerevisiae strain used was YPH499 (MATa ade 2-101 trp1-63 leu2-1 ura3-52 his3-200 lys2-801] [P. Hieter]). EPEC espG was subcloned between the BamHI and SalI sites of pEG(KG) (25). YPD (1% yeast extract, 2% peptone, 2% glucose) broth or agar was the general nonselective medium used for growing the yeast strains. Synthetic complete medium (SC) contained 0.17% yeast nitrogen base without amino acids, 0.5% ammonium sulfate, and 2% glucose and was supplemented with appropriate amino acids and nucleic acid bases. Galactose induction experiments in liquid medium were performed by growing cells in SC-Raf medium to log phase and then adding galactose to 2% for 6 to 8 h. Observation of actin in yeast cells with fluorescein isothiocyanate (FITC)-conjugated phalloidin (Sigma) was performed as previously described (18). Indirect immunofluorescence on yeast cells was performed as previously described (3). The following antibodies were used at the indicated dilutions: rabbit anti-glutathione S-transferase (GST; 1:500; Santa Cruz Biotechnology), rat anti-alpha tubulin (1:500; Serotec), mouse anti-actin (1:200; Sigma), indocarbocyanine (Cy3)-conjugated goat anti-rabbit immunoglobulin G (IgG; 1:500), FITC-conjugated goat anti-rat IgG (1:200), and FITC-conjugated goat anti-mouse (1:200). For phase-contrast, fluorescence microscopy and indirect immunofluorescence analyses, cells were examined with an Eclipse TE2000U microscope (Nikon). Digital images were acquired with Orca C4742-95-12ER charge-coupled device camera (Hamamatsu) and Aquacosmos Imaging Systems software.

RESULTS AND DISCUSSION

Role of EspG in bacterial virulence.

To assess the role of EspG in virulence, we constructed nonpolar deletion mutants of espG in both C. rodentium and EPEC. Both mutants behaved like their respective wt strains in LEE gene expression, type III secretion, and pedestal formation (6; data not shown). We infected two strains of mice (C57BL/6 and C3H/HeJ) mice with C. rodentium (wt and ΔespG) to assess mouse survival, bacterial colonization, and colonic hyperplasia. The more resistant C57BL/6 mice create a challenging intestinal environment for C. rodentium, allowing colonization defects to be assessed readily (31). In contrast, the environment within the susceptible C3H/HeJ mice is more permissive, permitting the assessment of mouse survival. Mutation of espG did not affect C. rodentium's ability to infect C3H/HeJ mice fatally between days 6 and 10 p.i. (data not shown).

We used C57BL/6 mice to assess colonization ability. Colons from mice infected with wt C. rodentium at day 6 p.i. were significantly heavier than those infected with the ΔespG strain (mg colon wt, 125 ± 8; ΔespG, 90 ± 3, P < 0.05), a finding indicative of colonic hyperplasia. Colonic hyperplasia was further confirmed by microscopic histological analysis (data not shown). Colon weights were not significantly different at day 10 p.i. The ΔespG strain also exhibited a significantly delayed ability to colonize C57BL/6 mice. Approximately 100-fold-fewer CFU were recovered from mice infected with ΔespG relative to wt at day 6 p.i. (CFU wt, [1.3 ± 0.7] × 10⁸; CFU ΔespG, [1.7 ± 0.8] × 10⁶; P < 0.05). No difference in colonization was seen at day 10 p.i. These data are similar to those observed for the REPEC espG mutant in the rabbit infection model (11) and suggest a role for EspG in early stages of host colonization.

Adherence and invasion.

Although EPEC is generally not thought to be an invasive pathogen, experiments with cultured cells suggest that the Map and Tir effectors may somewhat enhance invasion into nonphagocytic cells (16). We have also observed a low frequency of C. rodentium invasion into infected mouse epithelia by using electron microscopy (data not shown). To determine the role of EspG in adherence and invasion, we performed EPEC adherence and invasion assays with HeLa cells. No significant difference in adherence was observed between wt (11.5% ± 1.2%) and ΔespG (12.2% ± 1.3%) after 1.5 h. We assessed invasion rates after 3 h and observed a >10-fold decrease in the ability of ΔespG to invade HeLa cells (wt, 3.5% ± 0.1%; ΔespG, 0.2% ± 0.1%), a finding in agreement with a previous report (11).

In vivo localization of EspG.

Immunostaining of mouse colonic tissues infected by C. rodentium has been utilized previously to elucidate the location of translocated Tir in epithelial cells (7). We utilized this technology to determine whether EspG is translocated into the intestinal epithelium in vivo and to determine its intracellular localization. We inoculated orally C3H/HeJ mice with 2.5 × 10⁸ CFU of C. rodentium (wt or C. rodentium ΔespG) carrying a plasmid expressing C. rodentium espG-HA₂. After 4 days, we prepared intestinal epithelial tissue sections and immunostained against the HA epitope and detected expression of EspG in the host epithelium (Fig. 1). EspG was translocated and reached significant concentration in the apical surface of the host epithelium. We also stained against Tir to examine the relative locations of these two effector proteins (Fig. 1B). The distribution of Tir overlapped extensively with that of EspG. It remains unclear whether the apical surface localization of EspG reflects a simple proximity to the attached bacteria or whether it indicates the predominant site of activity of EspG in vivo. These data represent to our knowledge the first utilization of an epitope-tagged bacterial effector protein to determine host cell localization in vivo and represent the first example of simultaneous immunostaining of multiple effector proteins in an infected animal.

FIG. 1.

In vivo staining of translocated EspG in intestinal epithelial cells. Mouse tissue sections were collected after infection with C. rodentium expressing EspG-HA₂ and stained with an α-HA monoclonal antibody (red), α-Tir (green), or DAPI (blue). (A) Surrounding intestinal epithelium (magnification, ×630); (B) magnified view of an infected epithelial cell to illustrate effector protein localization.

EspG binds to mammalian tubulin.

Transformation of EPEC espG into S. flexneri ΔvirA complements the intracellular persistence phenotype, suggesting functional conservation of the biochemical activities of EspG and VirA in host cells (11). VirA binds to tubulin heterodimers to destabilize microtubules, triggering a signal transduction cascade to promote localized membrane ruffling and bacterial entry (34). We therefore sought to determine the mammalian protein(s) that interact with EspG after translocation into host cells. We screened a mouse spleen cDNA library (∼two million clones) with EPEC EspG as the bait protein and obtained two clones corresponding to amino acids 2 to 230 of mouse alpha-tubulin (three other clones were obtained and will be described elsewhere). A full-length human alpha-tubulin sequence was subcloned for directed bacterial two-hybrid assays. Plating assays on selective media demonstrated a 10-fold increase in the number of CFU, whereas quantitative β-galactosidase assays yielded a fourfold activation of the reporter gene cassette (data not shown). Thus, our genetic data indicate that EspG and alpha-tubulin form a stable complex inside a bacterium.

To confirm these genetic data with biochemical evidence of an EspG-tubulin interaction, we performed gel overlay assays. EPEC EspG was expressed and purified as an N-terminally His-tagged fusion protein. Tubulin heterodimers were used to overlay purified His-EspG. We detected specific binding to His-EspG but not to bovine serum albumin used as a negative control (Fig. 2A). We performed the converse experiment by first electrophorescing tubulin followed by overlay with His-EspG (Fig. 2B) and obtained similar results. These data indicate that the EspG is able to form a stable complex with tubulin in vitro in the absence of any cofactors.

FIG. 2.

EspG binds to mammalian tubulin. (A) bovine serum albumin (BSA) (left) or His-EspG (right) was electrophoresed by SDS-10% PAGE and transferred to nitrocellulose. Tubulin was overlaid and detected with an α-tubulin monoclonal antibody. (B) BSA (left) or tubulin (right) was electrophoresed on SDS-10% PAGE and transferred to nitrocellulose. His-EspG was overlaid and detected with an α-His monoclonal antibody.

EspG induces localized microtubule depolymerization and the formation of actin stress fibers.

After demonstrating through genetic and biochemical means that EspG binds tubulin, we assessed microtubule structure in infected epithelial cells. Given the intriguing homology and potential functional conservation between EspG and the VirA protein, we hypothesized that translocation of EspG might result in localized microtubule depolymerization inside host cells.

To analyze the effect of EspG on microtubule structure in infected host cells, we immunostained Caco-2 cells against tubulin after 4 h infection with either wt or ΔespG EPEC. In cells infected with wt EPEC, tubulin staining was diminished at the site of bacterial attachment, suggesting a localized destabilization of microtubules near the site of infection (Fig. 3A). In contrast, cells infected with the ΔespG strain displayed no diminished tubulin staining. To confirm the role of EspG in mediating microtubule disruption, we complemented the EPEC ΔespG mutant with pespG-HA₂, which expresses EspG fused to a hemagglutinin (HA) epitope. Host cells infected with EPEC ΔespG/pespG-HA₂ again displayed localized microtubule disruption, suggesting that EspG is important to this phenotype.

FIG. 3.

Translocated EspG induces localized microtubule depolymerization. (A) Assessment of local microtubule structure during infection. Caco-2 cells were infected with either EPEC wt (left), EPEC ΔespG (middle), or EPEC ΔespG/pespG-HA₂ (right) and stained with DAPI (blue) and an α-tubulin monoclonal antibody (red). Two representative cells are shown for each infection condition. (B) Distribution of EspG-HA₂ in infected HeLa cells. HeLa cells were infected with either EPEC wt/pespG-HA₂ (top) or EPEC ΔespG/pespG-HA₂ (bottom) and stained with an α-tubulin monoclonal antibody (red) and an α-HA monoclonal antibody (green). (C) EspG does not modulate cellular pools of either tubulin or actin. After infection, Caco-2 cells were lysed and 5 μg of protein was electrophoresed on SDS-10% PAGE and transferred to nitrocellulose. Tubulin and actin concentrations were assessed through immunoblotting.

We assessed the intracellular localization of EspG by infecting HeLa cells with an EPEC strain expressing EspG fused to an HA epitope (EspG-HA₂). EspG-HA₂ was found primarily at the cell periphery, a finding consistent with in vivo findings (Fig. 1) and was observed partially to colocalize with the distribution of tubulin (Fig. 3B). Colocalization with tubulin was independent of the presence of wt EspG, since expression profiles of tubulin in both EPEC wt/pespG-HA₂ and EPEC ΔespG/pespG-HA₂ were similar. The diminished microtubule concentration near the site of bacterial attachment was not due to a change in the total cellular pools of tubulin, since immunoblot analysis indicated that tubulin levels are unchanged, regardless of the expression of EspG (Fig. 3C).

The observed activity of EspG is reminiscent of VirA, which locally destabilizes microtubules near the site of bacterial attachment. Such a disruption triggers a Rac1-dependent signal transduction cascade to promote localized membrane ruffling and bacterial entry (34). It has been suggested that shortening of microtubules may stimulate Rho activity, possibly through the release of specific factors required for Rho stimulation at the plasma membrane (33). Thus, microtubule instability may be linked to actin cytoskeleton dynamics (34).

We therefore assessed the distribution of actin in infected epithelial cells (Fig. 4). We noted the formation of distinctive actin stress fibers in cells infected with wt, but at a reduced frequency in cells infected with ΔespG (Fig. 4A). Stress fiber formation was enhanced through complementation of ΔespG with pespG-HA₂. Total cellular actin pools were also unchanged among infection conditions (Fig. 3C).

FIG. 4.

EspG induces actin stress fiber formation. (A) Assessment of infected cells. Caco-2 cells were infected with either EPEC wt (left), EPEC ΔespG (middle), or EPEC ΔespG/pespG-HA₂ (right) and then stained with FITC-conjugated phalloidin. (B) Assessment of transfected cells. HeLa cells were transfected with 0.5 μg of plasmid expressing either an N-terminal FLAG tag to EspG (amino acids M1-T398), EspG (V220-Q332), or vector alone (pCMV-Tag2B) and then stained 36 h posttransfection with FITC-conjugated phalloidin.

We hypothesized that transfection of EspG into HeLa cells would also induce the formation of actin stress fibers. To test this idea, we transfected FLAG-tagged EspG into HeLa cells and visualized actin structure 36 h posttransfection. Cells transfected with FLAG-EspG displayed significant stress fiber formation (Fig. 4B, left), whereas cells transfected with vector alone did not (Fig. 4B, right). Yoshida et al. have shown previously that amino acids 224 to 315 of VirA are sufficient to bind tubulin (34). Given the intriguing link between microtubule stability and actin dynamics, we therefore tested whether the homologous amino acid sequence in EspG would be sufficient to induce actin stress fibers. Transfection of this region (V220-Q332) into HeLa cells also induced easily visualized actin stress fibers (Fig. 4B, middle).

Heterologous expression of EspG in S. cerevisiae disrupts yeast microtubule and actin dynamics.

We have developed Saccharomyces cerevisiae as a model to study EPEC effector proteins and previously observed that the inducible expression of EspG causes strong growth inhibition and activates the Slt2 MAPK pathway in yeast (27). To provide independent support of our data suggesting a link between disrupted microtubules and the actin cytoskeleton, we utilized the yeast model. We first examined the effect of EspG on yeast microtubule structure (Fig. 5). The spindle looked normal in shape in both mitotic GST- and GST-EspG-expressing cells. However, tubulin staining revealed that mitosis was uncoupled to budding in most cells expressing GST-EspG, since it was common to observe unbudded cells with a premitotic G2 spindle (Fig. 5B) or small-budded cells with two nuclei (Fig. 5C). These observations indicate that mitosis can take place but that nuclear segregation is erratic. The typical wt G1 conformation (Fig. 5A) of several astral microtubules emanating from the spindle pole body (SPB) was lost in cells expressing GST-EspG, with most G1 SPBs bearing only one short microtubule or even none. Instead, it was common to see unique long microtubules extended along cell surface, sometimes even spanning the whole-cell perimeter, and frequently detached from the SPB (Fig. 5D to E). These data imply that cytoplasmic microtubules lose their structure and function as a consequence of EspG and reveal a dramatic loss of coordination between bud development and nuclear division, processes that are tightly synchronized in yeast (2).

FIG. 5.

Expression of GST-EspG in budding yeast disrupts microtubule structure. (A) Control cells expressing GST, indicating typical G1 conformation of three astral microtubules emanating from the spindle pole body. (B to E) Abnormalities seen in cells expressing GST-EspG. Cells were stained with an α-GST monoclonal antibody (red), α-tubulin (green), or DAPI (blue).

We also examined the effect of EspG on the yeast actin cytoskeleton by staining fixed cells with fluorochrome-conjugated phalloidin. During bud development, actin patches normally accumulate at the growing bud (Fig. 6, left). However, cells with mislocalized actin cortical cytoskeletons were abundant when expressing EspG (Fig. 6, right). The observation that EspG-expressing cells accumulate small-budded cells is also consistent with a loss of ability to control cortical actin polarity.

FIG. 6.

Expression of GST-EspG in budding yeast disrupts the yeast actin cortical cytoskeleton. Characteristic actin distribution for cells expressing GST (left) and GST-EspG (right) as revealed by staining with FITC-conjugated phalloidin.

Together, our data surrounding heterologous expression of EspG in yeast suggest that EspG interacts with tubulin, as well as interferes with actin polymerization states through a potentially novel mechanism. It is possible that the defects in astral microtubule function we report in EspG-expressing cells are a consequence of impaired microtubule-capture sites due to loss of actin function. Indeed, Kar9-containing microtubule-capture sites at the cell cortex depend on the function of cortical actin (24). EspG may interfere with actin polarization cues, thus affecting both cortical actin and cytoplasmic microtubules.

In the present study we have shown that EspG plays a role in the ability of C. rodentium to colonize mice and to cause colonic hyperplasia. EspG was detected in apical regions of infected epithelial cells, where it may alter the host cytoskeleton to promote effective colonization. Genetic screens and gel overlay experiments indicate that EspG binds tubulin, while both immunofluorescence of infected tissue culture cells and heterologous expression of EspG in yeast demonstrate a disruption of native microtubule structure. Experiments that assessed actin structure also demonstrated an EspG-dependent induction of actin stress fibers in tissue culture cells and a loss of cortical actin polarity in yeast. It remains of interest to identify the host factor linking microtubule and actin dynamics in cells affected by EspG. It is interesting to speculate that EspG contributes to the ability of A/E pathogens to establish infection through a modulation of the host cytoskeleton involving transient microtubule destruction and actin polymerization in a manner akin to the Shigella flexneri VirA protein.

EspG may account for the observed decrease in abundance of Rac1, Rho GEF, and Rho GDI-α observed in infected Caco-2 cells (14). The phenotypes induced by EspG are reminiscent of VirA, which locally destabilizes microtubules near the site of bacterial attachment. Such a disruption triggers a Rac1-dependent signal transduction cascade to promote localized membrane ruffling and bacterial entry (34). It has been suggested that shortening of microtubules may stimulate Rho and Rac1, thus linking microtubule instability to actin dynamics (32). It is possible that EspG may modulate GEF-H1 signaling to regulate Rho activity (26).

ADDENDUM

While the present study was in review, Matsuzawa et al. published a manuscript that demonstrates that EPEC EspG and its homolog Orf3 induce actin stress fiber formation through a process involving localized microtubule destruction and the release of microtubule-bound GEF-H1 (21). That we are able to observe a role for EspG in modulating host microtubule and actin structures, even in the presence of native Orf3, may reflect differences in the bacterial strains and mammalian cell types utilized in each study.