Pore-forming Activity of the Escherichia coli Type III Secretion System Protein EspD*

Background: The membrane protein EspD is critical for pathogenic E. coli to inject virulence factors into host mammalian cells. Results: EspD inserts into membranes and forms an ∼2.5-nm pore. Conclusion: Pore assembly is dependent on anionic phospholipids and acidic pH. Significance: Elucidating the structural mechanisms of pore formation advances understanding of the T3SS function in EPEC and EHEC infections. Enterohemorrhagic Escherichia coli is a causative agent of gastrointestinal and diarrheal diseases. Pathogenesis associated with enterohemorrhagic E. coli involves direct delivery of virulence factors from the bacteria into epithelial cell cytosol via a syringe-like organelle known as the type III secretion system. The type III secretion system protein EspD is a critical factor required for formation of a translocation pore on the host cell membrane. Here, we show that recombinant EspD spontaneously integrates into large unilamellar vesicle (LUV) lipid bilayers; however, pore formation required incorporation of anionic phospholipids such as phosphatidylserine and an acidic pH. Leakage assays performed with fluorescent dextrans confirmed that EspD formed a structure with an inner diameter of ∼2.5 nm. Protease mapping indicated that the two transmembrane helical hairpin of EspD penetrated the lipid layer positioning the N- and C-terminal domains on the extralumenal surface of LUVs. Finally, a combination of glutaraldehyde cross-linking and rate zonal centrifugation suggested that EspD in LUV membranes forms an ∼280–320-kDa oligomeric structure consisting of ∼6–7 subunits.

Escherichia coli is a Gram-negative bacterium present in the intestinal microbiota of healthy humans (1,2). Certain strains of E. coli, however, have developed a sophisticated mechanism for adhesion and injection of bacterial virulence factors into gut epithelial cells. Enteropathogenic (EPEC) 3 and enterohemor-rhagic (EHEC) E. coli, such as the strain O157:H7, are pathogens responsible for gastrointestinal infections (3,4). When EPEC or EHEC cells come in close proximity to the host epithelial cells lining the intestinal tract, they express a set of proteins involved in the assembly of a structural organelle known as the type III secretory system (T3SS) (5). The T3SS is capable of translocating proteins directly from the bacteria into the host cell cytosolic compartment, an event that is critical for pathogenesis (6). Bacterial effector proteins injected into gut enterocytes manipulate a variety of cellular functions; however, the most striking phenotype associated with EHEC and EPEC is a remodeling of cytoskeleton and rearrangement of the actin filaments underneath the bacterial adhesion site resulting in a deformation of microvilli into a dysfunctional pedestal-like structure designated the site of attachment and effacement (7,8).
The proteins responsible for the attachment and effacement lesion are encoded on a pathogenicity island of ϳ35 kb referred to as the locus of enterocyte effacement (9,10). In EPEC, the locus of enterocyte effacement has been estimated to contain 41 open reading frames that encode effector molecules such as EspF, EspG, EspH, Map, regulatory proteins Ler, and GrlA/ GrlR (11)(12)(13), an outer membrane adhesin known as intimin and its receptor Tir (14), as well as the translocation apparatus that consists of the proteins EspA, EspB, and EspD (15)(16)(17)(18)(19)(20). Insertion of EspD into the cell membrane has been postulated to be an essential step in the formation of a translocation pore through which bacterial effector proteins are injected into the host cell (19).
In EPEC and EHEC, espB and espD have been genetically demonstrated to be important virulence factors as deletion of either of these genes impaired bacterial adhesion and infectivity (20 -26). Additionally, it has been suggested that EspB and EspD participate in pore formation, as loss of these proteins abolished pore assembly and injection of effector protein (19,21,23,27,28).
The EHEC bacterial strain EDL933 (O157:H7) espD encodes a 374-amino acid protein (39 kDa) predicted to contain two coiled-coil motifs located near the N (COIL I) and C (COIL II) termini, regions postulated to mediate EspD oligomerization ( Fig. 1) (29,30), and two transmembrane helices that form a hairpin structure that inserts into the host membrane. Although EspD is recognized to be a critical EPEC and EHEC virulence factor, little is known about the mechanism associated with membrane binding, oligomerization, and pore formation by this protein. Previous studies have demonstrated that the N-terminal region of EspD contains two amphipathic helices that mediate the initial anchoring of this protein to the lipid bilayer (29). Whether EspD alone in vivo is sufficient for translocon formation is not clear as previous studies detected EspB and EspD in plasma membranes (19). Here, we show that recombinant EspD, like native EspD secreted by EHEC, inserts into anionic lipid bilayers containing phosphatidylserine and forms a pore with an inner diameter of ϳ2.5 nm.

Experimental Procedures
Chemicals and Reagents-Restriction endonuclease DNAmodifying enzymes and cell culture reagents were obtained from Invitrogen. The pTYB2 vector and chitin affinity beads were purchased from New England Biolabs (Ipswich, MA). Electrophoresis and chromatography products were obtained from Bio-Rad, and clostripain was purchased from Promega (Madison, WI). Phospholipids were obtained from Avanti Polar Lipids, Inc. (Alabaster, AL), and chromatography grade sphingomyelin was obtained from Sigma. All other reagents were of the highest quality commercially available.
Purification of Recombinant EspD-The EHEC espD open reading frame was amplified from E. coli 0157:H7 EDL933 genomic DNA and cloned into NdeI and EcoRI restriction endonucleases in the pTYB2 plasmid (New England Biolabs) to generate the pTYB2-EspD expression vector used to express the EspD-chitin binding domain (CBD) fusion protein. For protein expression, the E. coli strain C41 (DE3) transformed with the pTYB2-espD construct was grown in Terrific Broth (1 liter) containing 100 g/ml ampicillin and 1% glucose at 37°C with vigorous shaking to an A 600 of 0.8, and then benzyl alcohol was added (10 mM final concentration) and the culture incubated at 20°C for 30 min prior to the addition of isopropyl thiogalactoside (0.5 mM). Cultures were incubated for 75 min at 37°C to induce EspD-CBD protein expression. The cell pellet was resuspended in 80 ml of Buffer A (10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , 500 mM NaCl, pH 7.2, 10% glycerol, containing a protease inhibitor mixture tablet (Roche Applied Science)), and cells were lysed by sonication. Clarified supernatants were applied to a chitin column (2.0 ml) equilibrated in Buffer B (10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , 500 mM NaCl, pH 7.2, 10% glycerol, 0.2% Triton X-100) and incubated with end-on-end mixing for 16 h at 4°C. The chitin column was washed with 400 ml of Buffer A, and the EspD fusion was cleaved with 50 mM dithiothreitol in Buffer B for 48 -72 h at 4°C. EspD was eluted, and the solution was made up to 1% Triton X-114. EspD was isolated in the detergent phase following phase separation per-formed at 37°C (31). The detergent phase containing EspD was collected by centrifugation at 18,000 ϫ g at 20°C for 10 min, and the Triton X-114 was removed by acetone precipitation. The purified protein was stored at Ϫ20°C as pellet and dissolved in 4.0 M urea in phosphate-buffered saline immediately prior to use. Protein concentrations were determined spectrophotometrically as described previously (32).
HeLa Cell Culture and Infection-HeLa cells (8 ϫ 10 5 cells/ well) grown to 80% confluence in 6-well tissue culture plates (Corning, Lowell MA) in DMEM with 10% heat-inactive bovine serum albumin were co-cultured with a 100-l aliquot of an overnight EPEC E2348/69 culture for 4 h at 37°C (29). The adherent cells were washed three times with phosphate-buffered saline (PBS) to remove bacteria, and the HeLa cells were gently scraped and resuspended in 1.0 ml of PBS containing protease inhibitor mixture and vigorously mixed using a vortex mixer to remove adherent bacteria. HeLa cells were lysed using a Dounce homogenizer (Kimble-Kontes, Vineland, NY), and crude membrane and cytosolic fractions were prepared by centrifugation at 14,000 rpm for 20 min at 4°C. The crude membrane preparation was extracted with 200 l of 1% sodium taurodeoxycholate for 30 min at 0°C. The insoluble material was removed, and the native EspD complex in the extract was analyzed by rate zonal ultracentrifugation.
Unilamellar Vesicle Preparation-The phospholipids dioleoylphosphatidylserine (DOPS), dioleoylphosphatidylcholine (DOPC), dioleoylphosphatidylethanolamine (DOPE), dioleoylphosphatidylglycerol (DOPG), palmityloleoylphosphatidylserine (POPS), palmityloleoylphosphatidylcholine (POPC), sphingomyelin (SM), cholesterol (Chl), or mixtures of these lipids dissolved in chloroform were evaporated using nitrogen gas to generate thin films. Residual solvent was removed by incubating films under reduced pressure for 16 h. The lipid films were dispersed in 40 mM sodium phosphate, 150 mM NaCl buffer, pH 7.2 (PBS), by vigorous mixing to form multilamellar vesicles that were incubated at 25°C for 1 h prior to extruding the mixture 30 times through a 200 nm polycarbonate filter using an extruder (Avanti Polar Lipids) to generate large unilamellar vesicles (LUVs). To prepare small unilamellar vesicles (SUVs), the multilamellar vesicles were sonicated with a probe sonicator on ice for 2-5 min until the lipid solution became clear. The resulting SUVs were annealed for 1 h at 37°C. SUVs were used in intrinsic fluorescence studies to minimize the degree of light scattering.
Membrane Binding Assay-LUVs (500 g) were incubated with recombinant EspD or espD mutant proteins (5 g) in 300 l of 40 mM sodium acetate, pH 4.5, 150 mM NaCl (SAS), or PBS, pH 7.2, for 15 min at 25°C and then diluted with 1.2 ml of 66% sucrose, 350 mM NaCl in PBS buffer. These preparations were overlaid with 3.0 ml of PBS, supplemented 350 mM NaCl and 35% sucrose and then 0.7 ml of PBS, pH 7.2. Samples were subjected to centrifugation at 28,000 rpm for 16 h at 4°C in a Beckman SW55 rotor. Fractions (0.75 ml) were collected from the top of the gradient, and proteins were precipitated with 12% trichloroacetic acid (TCA). Protein pellets were resuspended in SDS-PAGE sample buffer, and the distribution of EspD was examined by Western blot analysis.
To determine the apparent EspD membrane binding affinities, 667 M phospholipid in the form of SUV with composition SM/DOPC/DOPS (54:26:20), SM/DOPC/DOPS/Chl (44:24:12: 20), or SM/DOPC/DOPE/Chl (44:24:12:20) was mixed with increasing concentrations of EspD (46 -1330 nM) in 75 l of SAS or PBS buffer and incubated at 25°C for 10 min. The reaction mixture was made to 40% sucrose and transferred to a thick-walled 1.5-ml tube and then overlaid with 1.0 ml of 25% sucrose and 150 l of PBS. Samples were centrifuged at 49,000 rpm in a TLA100.3 rotor for 2.5 h at 4°C on a Beckman Coulter tabletop ultracentrifuge. Gradients were fractionated (250-l aliquots), and EspD was precipitated with 0.1% deoxycholate and 12% TCA. Protein pellets were dissolved in 25 l of 8.0 M urea. Aliquots of each fraction diluted in 100 l of distilled H 2 O/well were applied to a microtiter plate that was incubated for 16 h at 37°C uncovered to evaporate the distilled H 2 O and promote binding of EspD. Plates were blocked with 2% bovine serum albumin in PBS containing 0.1% Tween 20 (PBST) and then probed for 1 h at 37°C with anti-EspD primary antibody diluted 1:4000 in 2% BSA/PBST. Bound primary antibody was detected with a goat anti-rabbit horseradish peroxidase-conjugated secondary antibody, and plates were developed with the colorimetric substrate 3,3Ј,5,5Ј-tetramethylbenzidine.
Alkaline Carbonate and Triton X-100 Extraction-SM/ DOPC/DOPS/Chl (44:24:12:20) LUVs (500 g) loaded with EspD (5 g) were incubated at 25°C for 15 min in 40 mM sodium acetate, pH 4.5, 150 mM NaCl or PBS, pH 7.2, prior to purification of LUVs by sucrose density flotation centrifugation. LUVs harvested from the top of the density gradient were sedimented by centrifugation at 50,000 rpm for 40 min at 4°C. The LUV pellet was sequentially extracted with 1.0 ml of 500 mM NaCl, 100 mM Na 2 CO 3, pH 11.5, and 8.0 M urea for 30 min at 25°C. After each treatment the sample was separated into a supernatant and pellet fraction by centrifugation at 50,000 rpm for 40 min at 4°C. Proteins were precipitated with 12% TCA and the distribution of EspD in the pellet and supernatant was assessed by Western blot analysis.
To verify that recombinant EspD spontaneously inserted into mammalian cell plasma membrane, HeLa cells grown in a 24-well microculture plate to ϳ75% confluency were incubated with 26 g of recombinant EspD for 3 h in 1.0 ml of serum-free DMEM. The cells were washed with warm PBS to remove unbound protein, then scraped in 1.0 ml of ice-cold PBS containing a protease inhibitor mixture (Roche Applied Science), and lysed with a Dounce homogenizer, and crude membrane and cytosolic fractions were isolated by centrifugation at 14,000 rpm for 30 min at 4°C. The membrane pellet was washed with 1.0 ml of cold PBS and sequentially extracted with 500 mM NaCl, 100 mM Na 2 CO 3 , pH 11.5, and 8.0 M urea. Samples were fractionated at 50,000 rpm for 40 min at 4°C, and the distribution of EspD was determined by Western blot.
LUVs with SM/DOPC/DOPS/Chl were loaded with EspD at pH 4.5 or 7.2 and incubated for 15 min at 20°C. The pH in both samples was adjusted to 7.2 and then placed on ice for 15 min prior to extraction with 1% Triton X-100 for 30 min at 0°C prior to ultracentrifugation at 50,000 rpm for 40 min at 4°C in a Beckman-Coulter tabletop centrifuge equipped with a TLA100.3 rotor. The distribution of EspD in the supernatant and pellet fraction was assessed by Western blot analysis.
Dye Leakage Assays-Lipid thin films were resuspended 50 mM sulforhodamine B (SRB) in PBS and then extruded through a 200-nm polycarbonate filter membrane. LUVs loaded with SRB were purified by gel filtration chromatography (Sephadex G50, GE Healthcare). For leakage assays, LUVs (20 g of total phospholipid) diluted in SAS or PBS buffer were treated with EspD (0.2 g), and the increase in fluorescence intensity was measured at 585 nm (at an excitation of 565 nm). The total amount of entrapped SRB was determined with the addition of 0.2% Triton X-100. To determine the effect of electrostatic interactions on pore formation, dye leakage assays were performed in SAS and PBS buffer containing 0.5 or 1.0 M NaCl. All leakage assays were performed at 25°C.
Diameter of EspD Pore-The diameter of the EspD pores was assessed by leakage assays using carboxyfluorescein and fluorescein isothiocyanate (FITC)-conjugated dextrans with molecular masses of 4, 10, 40, or 70 kDa (Sigma) and Stokes radii of 1.8, 2.3, 3.2, 4.4 nm, respectively. Lipid films composed of SM/DOPC/DOPS/Chl were resuspended in PBS containing FITC dextran (5 mg/ml), and LUVs were prepared by extruding the multilamellar vesicle suspension through polycarbonate filters (200 nm). Lipid films composed of SM/DOPC/DOPS/Chl were resuspended in PBS containing FITC dextran (5 mg/ml), and LUVs were prepared by extruding the multilamellar vesicle suspension through polycarbonate filters (200 nm). FITC dextran-containing vesicles were purified by centrifugation at 50,000 rpm for 40 min at 4°C in a Beckman-Coulter tabletop ultracentrifuge using a TLA100.3 rotor, and LUVs were washed three times with 1.0 ml of PBS to remove free FITC dextran.
FITC dextran-loaded LUVs (200 g of phospholipid) were incubated with or without EspD (1 g) in 500 l of 20 mM sodium acetate, pH 4.5, 150 mM NaCl for 10 min, and the LUVs were sedimented by centrifugation at 50,000 rpm for 40 min at 4°C. Following centrifugation, the supernatant was removed, and the LUV pellet was dissolved in 500 l of 0.2% Triton X-100 in PBS. The fluorescence intensity of the pellet fraction was determined at an excitation of 492 nm and emission of 520 nm.
% leakage ϭ ͑F con Ϫ F Esp D͒/F con ϫ 100 where F con is total fluorescence in the untreated control LUVs, and F EspD is the total fluorescence in LUVs treatment with EspD (Equation 1).
Dual Quenching and Q Ratio-The interaction of EspD with lipid bilayers was examined by changes in the intrinsic fluorescence of espD mutants containing a single tryptophan (Trp) residue at positions Trp-47, Trp-178, or Trp-195. espD (5 g) was mixed with SM/DOPC/DOPS/Chl (44:24:12:20) SUVs (50 g) in 100 l of 20 mM sodium acetate, pH 4.5, or PBS and incubated at 25°C for 5 min prior to recording the emission spectra from 300 to 420 nm. The solvent accessibility of the single Trp residue was measured by fluorescence quenching experiments using 0 to 0.25 M acrylamide and recording emission spectra. mM NaCl, pH 4.5, or PBS, pH 7.2. Samples were incubated at 25°C for 5 min prior to recording the emission spectra. Fluorescence experiments were performed on a Varian Cary Eclipse spectrofluorometer at an excitation wavelength of 295 nm with excitation and emission slit widths of 5 nm. Emission spectra were recorded from 300 to 420 nm. All spectra were background corrected by subtracting spectra obtained for SUVs alone.
where F 0 is the fluorescence intensity in the absence of quencher; F acry is the corrected fluorescence intensity in the presence 0.25 M acrylamide, and F 10-DN is the corrected fluorescence intensity in the presence of 10 mol % 10-DN (Equation 2).
Cross-linking Experiments-Full-length EspD (10 g) or clostripain digested of EspD (10 g) was added to SM/DOPC/ DOPS/Chl (44:24:12:20) LUVs (1000 g) in PBS, pH 7.2, or 20 mM sodium acetate, 150 mM NaCl, pH 4.5, and incubated at 25°C for 15 min. Prior to the addition of 0.4 mM glutaraldehyde, the pH of both samples was adjusted to 7.2 with a 1.0 M phosphate buffer. Reaction mixtures were incubated at 25°C; aliquots were removed at various time points, and the reaction was terminated by the addition of 50 mM Tris, pH 8.0. The oligomerization of EspD was evaluated by Western blot analysis on a 6 or 10% SDS-polyacrylamide gel.
Protease Digests-Recombinant EspD (10 g) resuspended in 50 l of 50 mM sodium phosphate buffer containing 2.5 mM dithiothreitol and 1 mM CaCl 2 was treated with clostripain (50:1 mol ratio, Sigma) at 37°C. Aliquots were removed at 5, 30, 60, or 180 min and at 16 h for Western blot analysis and for dye leakage assays as described above using an EspD/phospholipid (EspD/PL) ratio of 1:2000 at pH 4.5. For membrane binding assays, the EspD limit digest product generated after a 16-h clostripain digest was incubated with 500 g of SM/DOPC/ DOPS/Chl (44:24:12:20) LUVs in 20 mM sodium acetate, pH 4.5, for 20 min (EspD/PL of 1:2000), and the membrane-bound protein was detected by Western blot following sucrose density flotation centrifugation as described above using sucrose solutions prepared in 40 mM sodium phosphate buffer, pH 7.2, containing 500 mM NaCl.
To examine the membrane topology, SM/DOPC/DOPS/Chl LUVs (500 g) were loaded into 5 g of EspD in PBS, pH 7.2 or pH 4.5, in SAS buffer and incubated for 15 min at 25°C, and the pH was adjusted to 7.2 with 1.0 M sodium phosphate. Each LUV preparation was treated with a 50:1 mol ratio of EspD/trypsin (Sigma) in the presence and absence of 0.5% Triton X-100 at 25°C. Aliquots were removed at 0, 5, 30, 60, or 180 min, and the reactions were terminated by acidification with 12% TCA. Precipitated proteins were analyzed by Western blot analysis.
Rate Zonal Centrifugation-EspD (5 g) was loaded onto SM/DOPC/DOPS/Chl (44:24:12:20) LUVs (500 g) by incubating at 25°C for 15 min in 40 mM sodium acetate, pH 4.5, 150 mM NaCl. LUVs were sedimented by centrifugation at 50,000 rpm for 1 h in a Beckman-Coulter tabletop ultracentrifuge. The supernatant was removed, and the pellet was resuspended in 150 l of PBS containing 1% sodium taurodeoxycholate and incubated at 25°C for 30 min to solubilize the lipid bilayer. The LUV extracts or extracts prepared from HeLa cells infected with EPEC E2348/69 bacteria were layered on a 10 -50% linear sucrose gradient prepared in PBS containing 0.2% sodium taurodeoxycholate, and complexes were resolved by centrifugation at 41,000 rpm at 6°C for 18 h using an SW55 rotor. The mass of the EspD complexes was determined using a standard protein mixture containing bovine serum albumin (20 g, 66 kDa), hemoglobin (50 g, 64 kDa), catalase (100 g, 240 kDa), and bovine IgM (50 g, 950 kDa). Gradients were fractionated, and proteins were precipitated with TCA and then analyzed by Western blot using anti-EspD antibodies or by Coomassie Blue-stained SDS-PAGE (standard protein).
Bioinformatic Analysis of EHEC EspD-The transmembrane helices, coiled-coil motifs, and amphipathic helices in EspD were predicted using the HMMTOP, COILS, and Heliquest algorithms. Isoelectric points and net charge on EspD was estimated using the Protein Calculator version 3.4 program. All dye experiments were performed at least in triplicate using three independent preparation of EspD.

Results
Structural Analysis of EspD-Bioinformatic analysis of the E. coli 0157/H7 EDL933 EspD revealed that this protein contains two putative transmembrane helices spanning residues 176 -200 (TMDI) and 227-251 (TMDII), the latter of which contains a number of serine residues (27,30,33). In addition, EspD contains two coiled-coil motifs situated between residues 138 -171 (COIL I) and 334 -370 (COIL II) ( Fig. 1) as predicted using the COILS algorithm (34). More importantly, fragments encompassing residues 1-171 or 330 -370 have been experimentally confirmed to form homodimers in solution (29,30). It was postulated that following membrane binding these two EspD domains remain exposed to the aqueous environment and form contacts that facilitate oligomerization. Collectively, these analyses suggest that the N-and C-terminal segments are important for pore formation by contributing to EspD oligomerization. EspD also contains two regions (residues 24 -40 and 66 -83) with a high propensity to forming amphipathic helices, elements important for EspD function on HeLa cells (29).
Purification of EspD-Initial attempts to produce large amounts of recombinant E. coli EHEC EspD containing either an N-or C-terminal hexahistidine affinity in E. coli were confounded by very low expression levels that were only detected by Western blot analysis using an anti-EspD-specific polyclonal antibody (29). To improve production, EspD was expressed as an N-terminal fusion protein linked to a CBD by a proteinsplicing intein motif. Although an ϳ95-kDa band corresponding to the EspD-CBD fusion protein was not apparent on Coomassie Blue-stained gels of E. coli C41 whole cell lysates following isopropyl ␤-D-thiogalactopyranoside induction ( Fig.  2A), Western blot analysis confirmed the presence of two immunoreactive proteins of ϳ40 and 95 kDa, which corresponded to the cleaved EspD protein and the intact EspD-CBD fusion construct (Fig. 2B). These results indicated that a portion of the fusion protein was undergoing protein splicing in vivo to generate the mature EspD that lacked N-or C-terminal affinity tags. The EspD-CBD fusion protein in the E. coli cytosolic fraction was purified on a chitin column, and the tagless form of recombinant EspD was recovered in the column eluate following a 48 -72-h treatment with DTT to promote protein splicing (Fig. 2). The recombinant EspD was further purified by phase separation using the detergent Triton X-114 (31). EspD was recovered in the detergent phase, and the contaminating proteins remained in the aqueous phase. Coomassie-stained SDSpolyacrylamide gels showed that this approach produced EspD preparations with an apparent purity of ϳ90% (Fig. 2).
Lipid Bilayer Binding Activity of EspD-A hallmark feature of native EspD is the insertion of this protein into the host cell membrane (19,27). To examine this event and to evaluate the phospholipids required for membrane binding, recombinant EspD was incubated with SUV composed of single phospholipids or lipid mixtures containing sphingomyelin, phosphatidylcholine, and cholesterol, which were selected to approximate the composition of the outer leaflet of the mammalian cell plasma membrane (35). SUVs were used for these initial experiments because multiple lipid bilayer composition could be rapidly generated by sonication. Sucrose density centrifugation experiments showed that at 25°C in pH 7.2 buffer EspD bound quantitatively to DOPC, DOPS, and SM SUVs and was recovered at the top of the gradient. In contrast, no EspD binding was observed with DOPE SUVs. In the absence of liposomes recombinant EspD remained at the bottom of the sucrose gradient (Fig. 3A). To assess the effect of lipid mixtures, EspD was mixed with SUVs composed of SM/DOPC/DOPE, SM/DOPC/Chl, SM/DOPC/DOPS/Chl, or SM/DOPC/DOPE/DOPS, and in all cases, near quantitative binding of EspD was detected (Fig. 3B). This suggested that membrane binding was not impacted by the presence of DOPE. Addition of 500 mM NaCl, an ionic strength that disrupts electrostatic membrane-protein interactions, was not sufficient to dissociate EspD from SUVs (data not shown), suggesting that this interaction was stabilized by nonpolar contacts with the hydrophobic core of the lipid bilayer. In contrast, no binding was observed with DOPE/Chl SUVs (Fig.  3B). To assess the effect of extravesicular pH on membrane binding, EspD was incubated with SM/DOPC/DOPS/Chl LUVs at pH 4.5 and 7.2 prior to flotation centrifugation. Under both conditions, comparable levels of EspD were recovered from the top of the gradient (Fig. 3C, fractions 1-3), indicating that the pH did not influence the membrane binding event.
To investigate the nature of the EspD-membrane interaction, SM/DOPC/DOPS/Chl LUVs loaded with EspD at pH 4.5 or 7.2 were sequentially extracted with 500 mM NaCl, alkaline carbonate, and 8.0 M urea. Regardless of the pH, treatment of LUVs with 500 mM NaCl or 100 mM sodium carbonate, pH 11.5, conditions that remove extrinsic or peripheral membrane proteins (36), showed that the bulk of EspD was recovered in the membrane pellet, which indicated that EspD binding was stabilized by hydrophobic rather than electrostatic contacts with the lipid bilayer (Fig. 4). Following each treatment, the bulk of EspD partitioned with the membrane pellet, a behavior consistent with EspD inserting into the lipid bilayer and forming contacts with the hydrophobic core. Extraction of EspD-loaded LUVs with 8.0 M urea showed that recombinant EspD predominantly partitioned with the membrane pellet further verifying that the EspD-membrane interaction was mediated by nonpolar contacts with the core of lipid bilayer (Fig. 4).
To ensure that recombinant EspD exhibited a similar interaction with mammalian cell membranes, HeLa cells were  loaded with EspD, and a crude membrane preparation was sequentially extracted with 500 mM NaCl, alkaline carbonate, and 8.0 M urea. For all extraction conditions, EspD was quantitatively recovered with the membrane fraction and recapitulated the interactions observed with the LUVs.
Interestingly, SM/DOPC/DOPS/Chl LUVs loaded with EspD at pH 4.5 or 7.2, treated with Triton X-100 at 0°C, and then subjected to high speed centrifugation revealed that EspD partitioned near quantitatively with a Triton X-100-insoluble pellet that likely contained detergent-resistant lipid microdomains due to the high concentration of sphingomyelin and cho-lesterol in the LUVs (Fig. 4B) (37). This is consistent with previous reports showing that EspD inserted into HeLa cell membranes was resistant to Triton X-100 extraction (27).
Dye Leakage Induced by EspD-The pore-forming activity of EspD was assessed by treating SM/DOPC/DOPS/Chl (44:24:12: 20) LUVs loaded with the pH-insensitive self-quenching fluorescence dye SRB. Addition of EspD to LUVs at pH 7.2 at an EspD/phospholipid (EspD/PL) ratio of 1:1250 triggered only a modest dye release of 13% (Fig. 5A). However, shifting the extravesicular pH to 4.5 induced a more robust leakage of 58%. These data suggested that an acidic pH was required to facilitate membrane penetration or oligomerization of EspD (Fig. 5A).
Leakages assays performed using extravesicular buffers with pH 3.5 to 7.2 (EspD/PL, 1:1250) showed that the dye release was minimal between pH 5.5 and 7.2 (Fig. 5B). However, at a more acidic pH, dye leakage increased rapidly and reached a plateau of ϳ70%, with the mid-point occurring at pH 4.7 (Fig. 5B). It should be noted that EspD has a predicted pI of ϳ5.4, and consequently this protein would carry ϳ11 positive charges at pH 4.7. Because negatively charged membranes have been shown to be important for pore formation in other systems (38 -40), we examined the requirement of anionic phospholipid for EspD pore formation. Leakage experiments with SM/DOPC/DOPE/ Chl (44:24:12:20) LUVs showed that replacement of DOPS with DOPE abolished dye leakage (Fig. 5B) and confirmed a role for anionic phospholipids in EspD pore formation. Flotation experiments showed that EspD efficiently bound to SM/ DOPC/DOPE/Chl LUVs at neutral and acidic pH. The requirement for negatively charged phospholipids was further sub-  with EspD at pH 4.5 or 7.2 and purified by sucrose density flotation centrifugation were sequentially extracted with 500 mM NaCl, 100 mM alkaline carbonate, pH 11.5, and 8.0 M urea in PBS and separated into a supernatant (S) or pellet (P) fraction. The interaction with HeLa cell membranes was assessed by incubating cells with recombinant EspD and then subjecting a HeLa crude membrane pellet to the sequential alkaline and urea extraction described above. B, LUVs loaded with EspD at pH 4.5 or 7.2 were extracted at 0°C with Triton X-100, and the soluble and insoluble fractions were separated by centrifugation as described above. EspD distribution was assessed by Western blot analysis using anti-EspD antisera.
stantiated by the finding that vigorous dye leakage, as a function of pH, was also observed when DOPS was substituted with DOPG (Fig. 5A). These data suggest a two-step mechanism consisting of a membrane binding event followed by a membrane penetration event that requires an anionic phospholipid and an acidic pH.

JOURNAL OF BIOLOGICAL CHEMISTRY 25585
DOPS increased dye leakage to 32%. Interestingly, incorporation of addition of DOPS into the membranes did not significantly increase SRB leakage from LUVs.
Previous studies implicated cholesterol as an important membrane component in EPEC and EHEC virulence (41,42). Consequently, we evaluated the requirement of cholesterol in EspD pore formation by replacing a portion of the DOPC (Y) in SM/DOPC/DOPS/Chl (45:Y:20:X) with increasing amounts of cholesterol (X). Interestingly, LUVs lacking cholesterol, but containing 20 mol % DOPS, showed ϳ30% dye leakage Fig. 5C. Augmenting the level of cholesterol in membranes to 20%, an amount typically found in mammalian cell plasma membranes, did not significantly increase dye release at pH 4.5 or 7.2. This result indicated that the presence of cholesterol in lipid bilayers was not a prerequisite for EspD pore formation.
To assess whether phosphatidylserine was necessary for the initial contact of EspD with the membrane surface or for penetration of EspD through the lipid bilayer, binding experiments and dye leakage assays were performed in the presence of high NaCl concentrations. Addition of EspD to LUVs diluted in buffer containing either 0.5 or 1.0 M NaCl followed by sucrose density centrifugation to resolve LUV bound from unbound EspD revealed that in the presence or absence of high salt EspD (at a EspD/PL mol ratio of Ͼ1:4000) was quantitatively recovered with LUVs at the top of the sucrose gradient at pH 4.5 or 7.2 (data not shown). These data suggest that electrostatic interactions are not essential for promoting the initial contact of EspD with the membrane. This was not unexpected because previous studies suggested that an N-terminal amphipathic helix facilitated initial integration into the outer leaflet of the bilayer (29).
To evaluate the impact of phospholipid composition on the EspD-membrane the interaction affinity, binding studies were performed by titrating SM/DOPC/DOPS, SM/DOPC/DOPS/ Chl, or SM/DOPC/DOPE/Chl SUVs with increasing concentrations of EspD at pH 4.5 or pH 7.2 and the level of EspD bound to SUV quantified by ELISA following sucrose density centrifugation to separate free EspD. Saturation kinetics was observed for all three types of SUV, which suggested that the presence or absence of DOPS, DOPE, or cholesterol did not significantly alter the initial binding of EspD to these lipid bilayers. Similar binding kinetics were observed when the initial binding reaction was performed at pH 4.5 or 7.2 (Fig. 5D). It is interesting to note, however, that a slightly lower B max value was obtained when the binding was performed at pH 7.2. Curve fitting of the data revealed a single binding isotherm with apparent dissociation constants ranging from ϳ67 to 132 nM for the EspDmembrane interaction. No dramatic difference in EspD binding affinity was observed with membranes lacking cholesterol ( Table 1).
The importance of the anionic membrane for pore formation was next examined by performing dye leakage assays in the presence of increasing concentrations of NaCl. Leakage assays performed at pH 4.5 with SM/DOPC/DOPS/Chl LUVs revealed that NaCl above ϳ0.4 M NaCl caused a rapid diminishment of SRB release to ϳ17%, levels that were comparable with the leakage induced by EspD at pH 7.2 in the absence or presence of NaCl (Fig. 5E). Control experiments performed with the addition of NaCl alone to SRB-loaded LUVs showed no significant dye release. These data, together with the flotation assays, suggest that the presence of anionic lipids in the bilayer is important for EspD penetration through the membrane or in driving formation of a functional pore.
Initial dye leakage assays were performed with phospholipids containing dioleoyl fatty acids. To more closely approximate the fatty acid composition of phospholipids in the mammalian cell plasma membrane, assays were repeated with LUV containing POPC and POPS. Addition of increasing concentrations of EspD to LUV composed of dioleoyl phospholipids at pH 4.5 resulted in robust release of SRB, which reached a plateau of ϳ60% at an EspD/PL mole ratio of 1:500 (Fig. 5F). Similar experiments performed with LUV composed of POPC and POPS resulted in a dye leakage of ϳ57% at a EspD/PL ratio of 1:500. The rate of SRB leakage was notably lower and suggested that the reduced fluidity of the membrane containing POPC and POPS lipids impact the kinetics of pore formation (Fig. 5F). Parallel experiments performed with DOPC/DOPS and POPC/ POPS LUVs at pH 7.2 showed a linear response that suggested that the accumulation of high levels of this protein on the lipid bilayer likely caused a disruption of the membrane integrity that may not involve pore assembly (Fig. 5F).
Orientation of EspD in Membranes-To investigate the topology of the EHEC EspD in SM/DOPC/DOPS/Chl LUV at pH 4.5 and 7.2, we exploited the three Trp residues located on amphipathic helix I (Trp-47) (29) or within the first transmembrane helix (Trp-178 and Trp-195) as intrinsic fluorescence probe to evaluate the interaction of these secondary structural elements with the lipid bilayer (Fig. 1). To dissect these proteinmembrane interactions, three espD variants containing only Trp-47, Trp-178, or Trp-195 were generated by introducing the mutations W178F,W195F, W47F,W195F, or W47F,W178F, respectively. Sucrose density flotation centrifugation experiments showed that at pH 7.2 all three espD mutants exhibited membrane binding activity similar to wild type EspD (Fig. 6A). Dye leakage assays demonstrated that at pH 7.2, like the wild type EspD, none of the espD mutants triggered notable dye leakage. However, at pH 4.5, all three espD mutants induced dye release comparable with the wild type EspD (Fig. 6B).
Fluorescence spectroscopy analysis of the espD Trp-47, Trp-178, or Trp-195 mutants in the absence of SUVs at pH 4.5 or 7.2 showed that espD mutants exhibited an emission max centered at ϳ346 nm (Fig. 6, C-E). The membrane binding of espD Trp-47 to SM/DOPC/DOPS/Chl SUVs at pH 4.5 and 7.2 resulted in a 4-nm blue shift in the emission max consistent with this residue inserting into a more nonpolar environment (Fig. 6C). A similar 3-nm blue shift in the emission max was observed when espD Trp-195 bound to SUVs at pH 4.5 (Fig.  6E). However, no notable change in the emission max of the Trp-178 mutant was detected in the presence of SUVs (Fig. 6D), which was diagnostic of Trp-178 localizing near the polar/nonpolar interface of the lipid bilayer. The binding of the three espD mutants to SUV membranes was accompanied by a notable decrease in the fluorescence intensity at pH 4.5 and 7.2. The only exception was espD Trp-47, which exhibited a slight increase in fluorescence intensity on binding SUVs at pH 4.5.
The espD mutants were used in dual quenching experiments that employed the quenching agents 10-DN and acrylamide (43) to evaluate the depth to which the Trp residues inserted into the lipid bilayer. Acrylamide preferentially quenches solvent-exposed or Trp residues located at the polar/nonpolar interface of the lipid bilayer. In contrast, 10-DN, a probe that embeds in the lipid bilayer core, quenches Trp residues that insert near the hydrophobic core of the membrane. The espD Trp-47, espD Trp-178, and espD Trp-195 mutants bound to SUVs at pH 4.5 showed a fluorescence quenching with acrylamide ranging from 0.82 to 1.22 (Table 2), values that are diagnostic of these Trp residues penetrating into the bilayer and becoming partially protected from acrylamide. Parallel experiments performed with SUVs containing 10-DN revealed that espD Trp-178 and espD Trp-195 were quenched to a greater extent than espD Trp-47 at pH 4.5 ( Table 2). The Q ratios calculated for espD Trp-47, espD Trp-178, and espD Trp-195 were 2.49, 1.1, and 0.65, respectively ( Table 2). The lower the value for the Q ratio, the deeper the Trp penetrates into the membrane. These values suggested that at pH 4.5, Trp-47 inserts into the membrane and localizes to the region near the polar/nonpolar interface, whereas the transmembrane helix containing amino acids Trp-178 and Trp-195 is predicted to insert perpendicularly into the membrane positioning these residues at different depths within the lipid bilayer.  (Table 2), consistent with Trp-195 penetrating deeply into the hydrophobic region of the outer leaflet of the membrane. This association with the outer leaflet was supported by the observation that EspD failed to form a functional pore at pH 7.2. In contrast, espD Trp-178 exhibited a significant degree of quenching with acrylamide and 10-DN, which implied that Trp-178 had considerable conformational flexibility that allowed this residue to shift in the membrane between a phospholipid headgroup interface and the nonpolar alkyl chain region of the outer leaflet (44). The large Q ratio obtained with espD Trp-47 at pH 7.2 was consistent with this residue penetrating deeply into the outer leaflet of the membrane. In contrast, Q ratios calculated for Trp-178 and Trp-195 predict that the transmembrane domain containing these aromatic residues inserts into the bilayer and positioned both residues at similar depths in the hydrophobic core (Table 2).
To further assess the orientation of the Trp residues when EspD is bound to the membrane, Stern-Volmer experiments were performed by titrating espD mutant-loaded SUVs loaded with increasing concentrations of acrylamide. Comparable Stern-Volmer constants (K SV ) obtained for espD Trp-47 at pH 4.5 and 7.2 ( Fig. 6F and Table 3) suggest that regardless of the environmental pH, Trp-47 inserts into the membrane near the phospholipid headgroups. At pH 4.5, espD Trp-178 and espD Trp-195 had K SV of ϳ5.4 M Ϫ1 (Fig. 6F and Table 3), values that were notably higher than K SV values for espD Trp-47. These differences suggest that Trp-178 and Trp-195 residues are less solvent-accessible and likely insert near the polar/nonpolar interface of the membrane. It is possible, however, that the espD oligomerization accompanying pore formation may shroud these aromatic residues from acrylamide quenching. However, at pH 7.2, the K SV values for Trp-178 and Trp-195 were dramatically higher (Table 3); this response indicated that the helix spanning residues 176 -200 inserted into the membrane in an orientation that was parallel to the lipid bilayer. This positions these Trp residues near the hydrophobic core and protected them from acrylamide quenching.
Membrane Topology of EspD-Studies with N-and C-terminal fragments of EspD showed that amphipathic domains and coiled-coil motifs in these regions (Fig. 1) were important for membrane binding and oligomerization on the host cell membrane (29,30). To evaluate the requirement of these regions for pore formation, EspD was treated with clostripain, an argininespecific endoproteinase that selectively removed residues 1-94 and 296 -373. Western blot analysis of digests showed a progressive truncation of EspD from the N and C termini resulting in a predicted protease-resistant fragment spanning residues 95-295 (Fig. 7A). Dye leakage assays performed at pH 4.5 revealed that this core fragment containing the transmembrane helix hairpin and COIL I (Fig. 7B), although capable of binding to LUV membranes (Fig. 7C), induced a dye leakage of ϳ14% compared with the 35% leakage obtained with the full-length protein. It is tempting to speculate that removal of the COIL II motif impairs the capacity of the espD(95-295) fragment to oligomerize and form a pore.
To further examine the EspD membrane topology, EspDloaded SM/DOPC/DOPS/Chl LUVs were treated with trypsin in the absence or presence of 0.5% Triton X-100. LUVs loaded at pH 4.5 and then digested with trypsin at pH 7.5 exhibited a notable resistance to proteolysis and required ϳ60 min for ϳ80% degradation of the parent protein (Fig. 7D). This resistance was attributed to EspD oligomerization and membrane insertion. Surprisingly, addition of Triton X-100 to LUVs loaded at pH 4.5 did not alter the degradation kinetics of EspD and suggested that the quaternary structure associated with pore formation contributed to the diminished rate of proteolysis (Fig. 7D). This conjecture is partially supported by the finding that EspD incubated in the presence of 0.5% Triton X-100 at pH 4.5 prior to shifting the pH to 7.5, for tryptic digestion, exhibited increased resistance to proteolytic cleavage and required ϳ15-30 min to achieve ϳ60% degradation. Alternatively, the association of the EspD complex with detergent-resistant microdomains at 0°C may protect EspD from tryptic digest (37). In contrast, LUVs loaded with EspD at pH 7.2 was highly susceptible to proteolysis and was completely degraded within 5 min, indicating that EspD was associated with the outer surface of the membrane and exposed to the protease (Fig. 7D). Solubilization of the latter LUVs with Triton X-100 did not alter this susceptibility to trypsin proteolysis.
Pore Formation by EspD-The effect of EspD concentration on pore formation was investigated by titrating SM/DOPC/ DOPS/Chl LUVs with increasing concentrations of EspD and monitoring SRB release. At pH 5.5 and 7.2, SRB release increased linearly to ϳ70% leakage at an EspD/PL ratio of  ϳ1:100 (Fig. 8A). This linear response reflects a slow rate of pore formation or a loss of membrane integrity due to the accumulation of EspD. In contrast, at pH 4.0 dye release was rapid and reached a plateau of ϳ65% at an EspD/PL ratio of 1:1000 (Fig. 8A), indicating that membrane penetration and oligomerization required for pore formation were probably facilitated by protonation of acidic residues. Addition of EspD to SM/DOPC/DOPS/Chl LUVs charged with the fluorescent probes carboxyfluorescein or 4-kDa FITC dextrans showed levels of 80 and 40% leakage, respectively. In contrast, dramatically lower levels of leakage were observed for the 10 -70-kDa FITC dextrans (Fig. 8B). This size-dependent release of these fluorescence probes was diagnostic of EspD forming a pore structure with an inner channel diameter of ϳ2.5 nm.
Oligomerization of EspD on model membranes at pH 4.5 or 7.2 was examined by treating SM/DOPC/DOPS/Chl LUVs loaded with EspD at pH 4.5 or 7.2 with 0.4 mM glutaraldehyde to cross-link and trap complexes. Using this approach, an ϳ80-kDa dimeric structure was rapidly detected after a 1.0-min incubation (Fig. 8C, upper panel). Extending the reaction time to 120 min failed to generate discrete higher ordered oligomers, likely due to glutaraldehyde cross-linking of DOPS to EspD; however, disperse EspD species migrating as a smear with a mass of ϳ150 kDa were detected by Western blots. Similar cross-linking experiments performed with LUVs initially loaded with EspD at pH 4.5 then shifted to pH 7.2, to facilitate the cross-linking reaction, showed that a dimeric structure was rapidly trapped. Under these conditions, additional high order species migrating as a smear of ϳ245 kDa were also observed (Fig. 8C, upper panel). EspD oligomerization on membranes is further supported by the disappearance of the EspD monomer as a function of cross-linking reaction time (Fig. 8C, lower  panel).
COIL I and II motifs (Fig. 1) have been proposed to mediate EspD-EspD interactions (29,30). To assess the role of the COIL I in EspD oligomerization, cross-linking experiments were performed with the EspD fragment encompassing residues 95-295 generated by clostripain digest (Fig. 7B). Glutaraldehyde treatment of this espD(95-295) loaded onto SM/DOPC/DOPS/Chl LUVs at pH 4.5 or 7.2 resulted in a rapid the entrapment of an ϳ40-kDa dimer. Prolonging the reaction time failed to trap additional higher order species (Fig. 8D).
Solubilization of LUVs loaded with recombinant EspD at pH 4.5 with the nondenaturing detergent taurodeoxycholate and then analyzing the complexes from EspD by rate zonal centrifugation on a linear sucrose gradient showed that the bulk of EspD was detected in fractions 1-3 where the monomeric form of recombinant EspD would migrate. However, a small portion of the EspD migrated with a mass of ϳ280 -320 kDa. It is unclear whether the abundance of EspD observed in the top fraction of the sucrose gradient is due to EspD that failed to oligomerize or a result of dissociation of the EspD complex following detergent extraction (Fig. 8E). Similar rate zonal centrifugation analysis of the EspD quaternary structure on LUVs loaded at pH 7.2 showed that EspD migrated nearly quantitatively as a monomeric species (Fig. 8E). These data suggest that membrane insertion is likely required for the assembly or stabilization of the EspD oligomeric structure.
To determine whether similar oligomeric structures were formed on host cell membranes during EPEC infection, membranes from HeLa cells infected with E. coli E2348/69 were extracted, and complexes formed by native EspD were examined. As with the recombinant EspD, both monomeric and oligomeric complexes with a mass of ϳ280 -320 kDa were also detected in the mammalian cell membrane (Fig. 8E). For both the recombinant and native EspD, it is unclear whether the monomeric species detected in these samples is due to membrane-associated EspD that failed to oligomerize or a result of complex dissociations following membrane extraction. The presence of the 280 -320-kDa complex, however, suggests that in membranes EspD assembles into structures containing 6 -7 subunits.

Discussion
In this study, we used recombinant EHEC EspD to examine the biochemical events associated with membrane binding, insertion, and pore formation in model membranes. Nascent EspD is predicted to associate CesD and CesD2, chaperones that are important for maintaining EspD in an unfolded state and preventing EspD from aggregation to the E. coli cytosolic compartment (45,46). On contact with epithelial cells, EspD is injected through the EspA needle complex (47) and is postulated to make an initial contact with the host membrane via an N-terminal amphipathic helix. This membrane-associated protein is thought to undergo oligomerization and subsequently insert into the membrane to form a translocation pore. To mimic this process in vitro, we used unfolded recombinant EspD to investigate the interaction of EspD with model LUV membranes.
A number of studies have suggested that cholesterol may be an important membrane component for EPEC and EHEC adherence, effector protein translocation, and pedestal formation (48 -51). However in these studies using LUVs with increasing concentrations of cholesterol, we demonstrated that this sterol was not required for EspD binding or pore formation. This contrasts previous observations with the T3SS proteins SipB and IpaB that not only exhibited a high binding affinity for cholesterol but this sterol was also required for efficient injection of T3SS effector proteins into mammalian cells (42,49,52).
Consequently, it is tempting to speculate that for EPEC and EHEC infections cholesterol may be required for the function of T3SS effector proteins that interact with membranes (41,51).
Membrane binding revealed that, with the exception of phosphatidylethanolamine, EspD spontaneously bound to unilamellar vesicles containing single dioleoyl or palmitoyloleoyl phospholipids or lipid mixtures containing sphingomyelin. Biophysical analysis confirmed that EspD-membrane interaction is stabilized by contacts with the hydrophobic core of the lipid bilayer. Surprisingly, although EspD bound to SM/DOPC/Chol vesicles, a composition resembling the outer leaflet of the mammalian cell plasma membrane (35), it failed to induce pore formation unless the anionic lipids phosphatidylserine or phosphatidylglycerol were incorporated into the FIGURE 8. Characterization of the EspD pore. A, effect of pH and EspD concentration on pore formation was examined by titrating SRB SM/DOPC/DOPS/Chl LUVs with increasing concentrations of EspD at pH 4.0, 5.5, and 7.2, and the percent dye release was plotted as a function of the EspD/phospholipid ratio. B, diameter of the EspD pore formed at pH 4.5 was determined with LUVs loaded with carboxyfluorescein or FITC labeled 4 -70-kDa dextrans. The oligomeric state of EspD on LUVs at pH 4.5 or 7.2 was assessed by glutaraldehyde cross-linking, and complexes were examined by Western blot on a 6% (upper panel) and 10% (lower panel) SDS-polyacrylamide gel (C). D, clostripain fragment of EspD encompassing residues 95-295 were glutaraldehyde cross-linked after loading protein onto LUVs at pH 4.5 or pH 7.2. E, complexes formed by recombinant and native EspD on LUV and HeLa cell membranes were examined by rate zonal centrifugation on a linear sucrose gradient following detergent extraction of complexes.
bilayer. This prerequisite suggests that phospholipid headgroups may contribute to the folding or penetration of EspD across the membrane (38,53,54). A similar requirement for negatively charged phospholipids has been noted for the action of the bacterial toxin colicin Ia (55) and the T3SS proteins PopB/D, IpaB, and SipB (38,39,52,54,56). Under homeostatic conditions, the plasma membrane exhibits an asymmetric phospholipid distribution with phosphatidylserine and phosphatidylethanolamine being located predominantly on inner leaflet of the membrane. However, contact of EPEC or EHEC with the epithelial cell mediated by the bundle forming pili causes an increased translocation of phosphatidylserine to the host cell surface (41,(57)(58)(59). In this study, we show that a modest amount of phosphatidylserine (5 mol %) was sufficient to facilitate insertion of the transmembrane helix hairpin resulting in pore formation by EHEC EspD.
The N-terminal amphipathic regions of EspD were shown to be important for infectivity and pedestal formation (29). Intrinsic fluorescence techniques confirmed that the helix containing Trp-47 (Fig. 1, Amph I) inserts into the polar/nonpolar membrane interface providing an initial contact that anchors EspD to the host membrane as it emerges from the T3SS needle complex. Fluorescence dual quenching and proteolytic mapping studies further support a model where at pH 7.2 the amphipathic and transmembrane helices bind the membrane by inserting in a parallel configuration into the outer leaflet of the lipid bilayer (Fig. 9A). However, this association is not sufficient to promote pore formation as indicated by dye and FITC dextran release assays. A similar activity has been reported for SipB and IpaB that spontaneously insert into host membranes but failed to induce hemolysis or the assembly of a functional pore (42,52,56). In contrast, glutaraldehyde cross-linking and rate zonal centrifugation experiments showed that EspD, which formed functional pores on LUVs at pH 4.5 or were isolated from infected HeLa cell membranes, formed an oligomeric complex of ϳ280 -320 kDa estimated to contain 6 -7 EspD subunits. The arrangement of 12-14 transmembrane helices in a barrel configuration (60) would be sufficient to form a pore with an inner diameter of ϳ2.5 nm as measured for EspD, a diameter comparable with the size of the central channel of the needle apparatus formed by EspA (61). Surprisingly quaternary structure analysis revealed that EspD bound to LUV membranes at pH 7.2 was monomeric. However, it is not possible to discount the possibility that at the membrane surface EspD may form transient higher order structures that, following penetration of the lipid bilayer, would be stabilized by the packing of the transmembrane domains.
EspD that inserts into membranes at neutral pH forms oligomeric structures that are stabilized by COIL I-COIL I and COIL II-COIL II contacts (29,30,62) that are formed between adjacent EspD subunits. These oligomeric complexes form a pre-pore-like structure (Fig. 9A). This model is supported by cross-linking showing that espD(95-295) formed predominantly dimers stabilized by COIL I-COIL I contacts (29) and despite retaining the transmembrane helix hairpin had impaired dye release activity likely due to the compromised capacity for this fragment to stable oligomers as a result of the loss of the COIL II domain.
Formation of a functional pore was dependent on the incorporation of an anionic phospholipid into membranes and an acidic pH, conditions that drive the penetration of helices of channel-forming toxins and antimicrobial peptides into the lipid bilayer (39,40,54,(63)(64)(65)(66). It is possible the insertion of EspD proceeds via a toroidal pore-forming mechanism that is known to be influenced by lipid composition (67) and may explain the requirement of anionic phospholipids for pore formation. It is conjectured that under acidic conditions the key amino acids on EspD become protonated, which lowers the thermodynamic barrier associated with insertion of charged residues through the hydrophobic core of the lipid bilayer (68,69). It is interesting to note that in the presence of high salt the dye release from LUVs was dramatically diminished, which FIGURE 9. Model of EspD membrane interaction. A, at pH 7.2 the transmembrane helices and the N-terminal amphipathic region of EspD are postulated to insert into the outer leaflet of the lipid bilayer in a parallel orientation with the coiled-coil motifs (COIL I and COIL II) remaining solvent-exposed and mediate EspD oligomerization through COIL I-COIL I and COIL II-COIL II interactions to form a pre-pore structure. B, conformational changes triggered by an acidic pH are postulated to induce structural rearrangements that involve the penetration of the hairpin loop and formation of a pore resulting in a re-orientation of the transmembrane domains in a configuration that is perpendicular to the lipid bilayer. C, multiple sequence alignments showing the distribution of acidic residues in the hairpin loop region for the EspD homologues from a Salmonella, Pseudomonas, Yersinia, and Shigella.
suggests that an electrostatic interaction between EspD and negatively charged lipids may be important for pore formation. This is further supported by binding studies suggesting that the interaction affinity of EspD with membranes was not significantly altered by the presence of phosphatidylserine or cholesterol in the lipid bilayer. The absence of the former phospholipid did impair pore formation. Because the N and C termini of EspD are predicted to remain on the extracellular surface of the plasma membrane, the conversion of the EspD pre-pore to pore structure requires the translocation of the transmembrane hairpin loop spanning residues 200 -227 (Fig. 9B) into the cytosolic compartment of mammalian cells. Dye leakage assays revealed that EspD pore formation was dramatically enhanced at a pH below 4.7, a value that approximates the pK a value for glutamate and aspartate ionization. Interestingly, the EHEC EspD hairpin loop contains two acidic amino acids, Asp-211 and Glu-215 (Fig. 9C), that would require protonation for penetration through the membrane that results in the re-orientation of the transmembrane helices from a parallel to perpendicular configuration with the membrane and packing of the helices from 6 to 7 EspD subunits to form a functional pore (Fig.  9B). Despite notable variations in length, the hairpin loops connecting the transmembrane domains of YopB, PopB, SipB, and IpaB all contain a number of acidic residues (Fig. 9C). Taken together with the observation that these translocator proteins display increased pore-forming activity at low pH, it is tempting to speculate that these charged loop residues play a pivotal role in pore assembly. However, this hypothesis needs to be validated by site-directed mutagenesis studies.
In the intestinal tract, it is conceivable that a localized acidic environment required for EspD membrane insertion may be generated by the following: (i) the metabolism of pathogenic E. coli microcolonies attached to epithelial cells (70); (ii) upregulating the activity of the sodium-hydrogen exchanger 2 antiporter, which may cause a decrease in the cell surface pH (70,71), and (iii) down-regulation of the monocarboxylate transporter 1 activity, which would result in the extracellular accumulation of short chain fatty acid metabolites and a decreased pH at the apical surface of the gut epithelial cells (72). It is also possible that contacts with EspA and EspB may be instrumental in driving EspD membrane insertion and pore assembly (28,30).