The Coiled-coil Domain of EspA Is Essential for the Assembly of the Type III Secretion Translocon on the Surface of EnteropathogenicEscherichia coli *

Enteropathogenic E. coli (EPEC) utilize a type III secretion system to deliver virulence-associated effector proteins to the host cell. Four proteins, EspA, EspB, EspD, and Tir, which are integral to the formation of characteristic “attaching and effacing” (A/E) intestinal lesions, are known to be exported via the EPEC type III secretion system. Recent work demonstrated that EspA is a major component of a filamentous structure, elaborated on the surface of EPEC, which is required for translocation of EspB and Tir. The carboxyl terminus of EspA is predicted to comprise an α-helical region, which demonstrates heptad periodicity whereby positions a and d in the heptad repeat unitabcdefg are occupied by hydrophobic residues, indicating a propensity for coiled-coil interactions. Here we demonstrate multimeric EspA isoforms in EPEC culture supernatants and EspA:EspA interaction on solid phase. Non-conservative amino acid substitution of specific EspA heptad residues generated EPEC mutants defective in filament assembly but which retained the ability to induce A/E lesions; additional mutation totally abolished EspA filament assembly and A/E lesion formation. These results demonstrate a similarity to flagellar biosynthesis and indicate that the coiled-coil domain of EspA is required for assembly of the EspA filament-associated type III secretion translocon.

Most bacterial virulence-associated determinants are either surface located or are secreted from the bacterium. There are only a limited number of ways by which Gram-negative bacteria can transport proteins across their unique double membrane (reviewed in Ref. 1). The type III secretion system, found in many Gram-negative pathogens (reviewed in Ref. 2), is responsible for delivery of virulence-associated factors involved in subversion of the host-cell signal transduction pathways required for bacterial adhesion, invasion, and disease. A particularly good example of bacterial pathogens that employ a multi-stage infection strategy involving a type III secretion system is provided by enteropathogenic Escherichia coli (EPEC) 1 and enterohemorrhagic E. coli (EHEC) (reviewed in Ref. 3). EPEC is an established etiological agent of human diarrhea, and EHEC is an emerging food borne cause of acute gastro-enteritis and hemorrhagic colitis (reviewed in Ref. 4). Subversion of epithelial cell function by EPEC and EHEC leads to the formation of distinctive "attaching and effacing" (A/E) lesions, characterized by localized destruction (effacement) of brush border microvilli, intimate attachment of the bacillus to the host cell membrane, and the formation of an actin-rich underlying pedestal-like structure in the host cell (reviewed in Ref. 3). All the genes necessary for the A/E effect map to a pathogenicity island termed the locus for enterocyte effacement or LEE (5), which includes structural components of the secretion apparatus (6), the adhesion molecule intimin (7,8), secreted proteins EspA, EspB, EspD (reviewed in Ref. 3), and Tir (9), and their respective chaperones (10,11).
The type III secretion apparatus comprises approximately 20 mainly inner membrane-associated proteins, which demonstrate broad functional conservation across bacterial species (2,12). Additionally, many of these components demonstrate similarity to proteins involved in bacterial flagellar biosynthesis (2). EspA is secreted via the type III secretion apparatus and is required for the translocation of EspB and Tir into the host cell cytosol (9,13,14). It has recently been identified as a major component of large extracellular filamentous structures, which appear on the surface of EPEC at an early stage of infection prior to intimate attachment (14).
The assembly of a number of eukaryotic proteins into filamentous structures is based upon interactions between dimeric coiled-coil domains (15), and similarly predicted coiled-coil segments lie within regions of functional significance in a number of bacterial proteins. Coiled-coil segments have been predicted to occur in many proteins associated with type III secretion systems (16), where they have a potential role in mediating the formation of large multi-protein complexes or more discrete interactions that may facilitate protein translocation. Notably, bacterial flagellins contain coiled-coil domains at the amino and carboxyl termini of the protein corresponding to regions deemed important in the polymerization of flagellin into the flagellar filament (17,18). EspA demonstrates discrete sequence similarity to flagellins in the carboxyl-terminal region of the protein which is predicted with high probability to adopt a coiled-coil conformation (16,19).
The similarity between type III secretion components and proteins of the flagellar export machinery, particularly in regions predicted to form coiled-coils prompted us to investigate the importance of the predicted coiled-coil domain of EspA for the assembly and function of the EspA filaments. In this study we provide evidence, based on a rational site-directed mutagenesis approach, that mutation of specific residues predicted to be critical in coiled-coil conformation have dramatic consequences for the assembly of the EspA filament and subversion of host cell signal transduction pathways leading to A/E lesion formation. Considering the similarities between different type III secretion systems, coiled-coil interactions may represent a common feature of their function.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Plasmids-The bacterial strains used in this study were generated by vector complementation of the espA gene in the espA mutant strain UMD872 (20). Complementation was achieved by transformation of plasmid pMSD2 containing the functional EPEC espA gene (20). In addition, strains UMD864 (espB Ϫ ), UMD870 (espD Ϫ ), and the prototype (parent) EPEC E2348/69 were used. Mutant strains were generated by site directed mutagenesis of pMSD2. Bacteria were grown to stationary phase at 37°C in L-broth, or Dulbecco's modified Eagle's medium (DMEM), with the addition of chloramphenicol to a final concentration of 30 g/ml as appropriate.
Sequence Analysis-Coiled-coil predictions were carried out using a revised version of the computer program described by Lupas et al. (21) available via the World Wide Web. Predictions were based on a window size of 28 residues and weighting the algorithm in favor of hydrophobic residues at positions a and d of the heptad repeat.
Construction of GST-EspA Fusion Protein and Solid Phase Binding Assay-For analysis of EspA monomeric interactions using GST-EspA fusion protein, a BamHI-flanked espA polymerase chain reaction fragment (primer pair; 5Ј-CGGGATCCATGGATACATCAACTAC-3Ј and 5Ј-CGGGATCCTTATTTACCAAGGGATATTCC-3Ј, temperature cycling as follows; 1 cycle of 95°C, 5 min, then 25 cycles of 95°C, 30 s, 58°C 45 s, 72°C 45 s, with a final extension step of 72°C, 5 min) was cloned into the BamHI site of the GST gene fusion vector pGEX-2T, creating a translational fusion with glutathione S-transferase (generating plasmid pICC34). The GST-EspA fusion was produced in XL1-Blue by induction of a 2-h log phase culture with 1 mM isopropyl-1-thio-␤-D-galactopyranoside and incubating for another 4 h at 37°C. GST or GST-EspA was purified from 10 ml of clarified sonicate by the addition of 100 l (50% slurry) of glutathione-Sepharose 4B beads (Amersham Pharmacia Biotech). Following 30 min of incubation, the beads were washed three times in PBS, and the fusion eluted by the addition of a total of 500 l of 10 mM reduced glutathione in 50 mM Tris -HCl, pH 8.0.
EspA interactions were subsequently assayed by ELISA as follows; ELISA plates were coated with 25 g/ml purified His-EspA (12) in carbonate/bicarbonate buffer overnight at 4°C. Wells were blocked, then incubated at 37°C for 2 h with three-fold serial dilutions of GST-EspA, GST, or GST-EspA with the addition of 5 g of His-EspA in PBS-Tween (0.05%). Plates were washed, then probed with anti-GST antibody (Amersham Pharmacia Biotech, 1/500 dilution, 1 h) followed by alkaline phosphatase-conjugated anti-goat antiserum (Sigma, 1/1000, 1 h). The reaction was visualized by addition of p-nitrophenyl phosphate in 100 mM Tris-HCl buffer, pH 9.5, and the final optical density measured at 405 nm. The mean optical density from five separate experiments was used for graphical representation.
Site-directed Mutagenesis-Site-directed mutagenesis of the espA gene was performed using the QuickChange site-directed mutagenesis kit (Stratagene) following manufacturer's instructions, using doublestranded pMSD2 or subsequently mutated vector as template. Complimentary mutagenesis oligonucleotide pairs incorporating single amino acid substitutions are as follows.
Mutated plasmid containing staggered nicks was generated by extension of primers annealed to opposite strands of the denatured plasmid by temperature cycling (1 cycle of 95°C, 30 s, then 16 cycles of 95°C, 30 s, 55°C, 1 min, 68°C, 18 min) in the presence of the high fidelity Pfu DNA polymerase. Synthesized DNA containing the desired mutation was selected from the original DNA template by incubation with DpnI at 37°C for 1 h. Nicks in the plasmid were repaired following transformation of 1 l of the synthesized products into competent E. coli XL1-Blue cells. Chloramphenicol-resistant transformants were randomly selected and inoculated to overnight L-broth cultures for preparation of plasmid Mini-preps (Qiagen). Correct incorporation of each mutation was monitored by DNA sequencing using an automated DNA sequencer (ABI 377). Mutated plasmid was transformed to competent UMD872 cells (espA Ϫ ) for analysis of phenotypic effects.
Polyacryamide Gel Electrophoresis and Western Blotting-Proteins secreted in DMEM supernatants (EspA, EspB) or contained in whole cell lysates (intimin) from induced cultures were separated on 12% reducing or non-reducing polyacrylamide gels. Proteins were then transferred to nitrocellulose membranes, which were subsequently blocked for 1 h by the addition of 5% bovine serum albumin in PBS/ Tween 20 (0.05%) as described (22). Membranes were variously probed for 2 h with anti-EspA (14), anti-EspB (14), or anti-intimin (22) antisera all at 1:1000 dilution in PBS/Tween 20 (0.05%). Following a standard wash step, blots were incubated for 2 h with anti-rabbit alkaline phosphatase-conjugated secondary antibody at a dilution of 1:5000 in PBS/ Tween 20 (0.05%). The immunoblots were subsequently developed by the addition of one buffered substrate tablet (Sigma) containing 0.15 mg/ml 5-bromo-4-chloro-3-indolyl phosphate and 0.30 mg/ml nitro blue tetrazolium dissolved in 10 ml of distilled water.
Fluorescence Actin Staining (FAS) and Detection of EspA Filaments-The ability of EspA derivative strains to induce A/E lesion formation was assessed using the FAS test (23), and detection of EPECassociated EspA filaments with anti-EspA polyclonal antiserum was performed according to Knutton et al. (14).

Identification of Multimeric EspA in Supernatants of EPEC
Cultures-We have previously suggested that EspA polymerization is the structural basis for the formation of EspA filaments, which, similar to flagella, are hollow cylindrical structures assembled in a manner dependent on a type III secretion system (3,19). In this study we sought to investigate the basis for EspA filament assembly.
We analyzed Western blots of secreted EPEC proteins, separated under non-denaturing conditions. This provided direct evidence for polymerized forms of EspA in culture supernatants of the prototype EPEC strain E2348/69 (Fig. 1, lane 4). No EspA was detected in culture supernatants of UMD872 harboring a deletion mutation in the espA gene (Fig. 1, lane 3), but multimerization was restored in UMD872 complemented with FIG. 1. EspA multimerization in the supernatants of EPEC cultures. Secreted proteins from mutant EPEC strains UMD870 (espD Ϫ ), UMD864 (espB Ϫ ), UMD872 (espA Ϫ ), and the prototype wild type strain E2348/69 were separated in 12% native gels, blotted, and probed for EspA (lanes 1-4, respectively). The shifted bands imply multimeric forms of EspA, however, due to the non-gradient nature of the gel it cannot be determined which specific multimeric forms are represented. Significantly, EspD and EspB mutant strains demonstrate the same laddered profiles as the wild type EPEC strain.
pMSD2 containing a cloned espA gene (results not shown). These results demonstrate that EspA is a major component of the EspA filament assembly, and as such, monomers would be expected to possess structural regions that would mediate hetero-and/or homo-oligomeric interactions. Importantly, multimerization of EspA was not affected in the background of espD and espB mutant strains and appeared identical to the wild type strain E2348/69 (Fig. 1, lanes 1 and 2, respectively).
Solid-phase EspA-EspA Binding Assays-As a preliminary demonstration of EspA:EspA interactions, we assessed the ability of purified EspA fusions to associate in an ELISA based assay. Purified His-tagged EspA was used to coat wells of a 96-well plate, then serial dilutions of purified GST-EspA were added. GST-EspA immobilized in wells by association with His-EspA was detected by anti-GST antibody followed by alkaline phosphatase-conjugated anti-goat antiserum. GST-EspA demonstrated a dose dependent association with His-EspA (Fig. 2). Additionally, GST-EspA:His-EspA interaction could be competitively inhibited by the addition of 5 g of soluble His-EspA in tandem with the GST-EspA fusion protein. Binding of GST alone by His-EspA was not evident. These results provide a direct demonstration of EspA interactions by a solid phase assay.
Site-directed Mutagenesis of Heptad Repeat Residues within the Carboxyl-terminal Coiled-coil Motif of EspA-We have previously shown that coiled-coil motifs are a common feature of many proteins secreted by type III secretion systems (16). In EspA, six heptad repeats comprising the motif aXXdXXX (where a and d represent hydrophobic residue and X any residue) can be identified in the carboxyl terminus of the EspA protein indicating a propensity of this region to adopt a coiledcoil conformation (Fig. 3A). When analyzed by the COILS algorithm (21), these repeats are predicted to form a coiled-coil with a probability of 98% (Table I). Amino acids in the heptad repeat of coiled-coils are assigned positions a-b-c-d-e-f-g, where the a and d residues are largely hydrophobic and form the hydrophobic core in an association between one or more helices (24). We aimed to introduce non-conservative substitutions at position a which are deemed important for the coiled-coil conformation. Non-conservative substitution of single, double, and triple position a residues in selected repeats (Fig. 3B) reduced the probability of coiled-coil formation by the values listed in Table I.
The EspA derivatives were constructed by site-directed mutagenesis using plasmid pMSD2. Mutant derivatives were introduced into UMD872, and biological functions of the mutated EspA were characterized in terms of Esp protein secretion, EspA multimerization, EspA filament assembly, and the ability to support A/E lesion formation.
Mutations in the Coiled-coil Domain of EspA Do Not Affect Protein Expression and Secretion-Western blots of bacterial supernatants and whole cell extracts were used to confirm that the mutations introduced into the coiled-coil domain of EspA did not affect LEE-encoded protein expression or secretion. All of the single and double espA mutants secreted EspA at overall levels comparable to wild type in both L-broth and DMEM cultures (Fig. 4A), although some culture to culture variation between experiments was observed (Fig. 4A). In contrast, the triple espA mutant UMD872(pICC24) produced an insoluble EspA that was only detected in the bacterial pellets and migrated with a higher molecular weight compared with the native EspA (data not shown). Consequently, the triple EspA mutant was not investigated further.  Mutagenesis of espA had no significant effect on secretion of EspB and although, similarly to EspA, some variations between experiments was observed, similar levels of secreted EspB were detected in culture supernatants of all strains (Fig.  4B); expression of intimin, an outer membrane EPEC adhesion molecule, by all mutant strains was similarly unaffected (data not shown). These results show that the mutations in the coiled-coil domain of EspA do not have pleotropic effects on expression or secretion of LEE-encoded virulence genes, including EspA itself.
Mutations in the Coiled-coil Domain Affect Multimerization of EspA-An indication of the impact of position a substitutions on the ability of EspA to polymerize into higher structures was obtained by analysis of Western blots of DMEM-secreted proteins separated under non-denaturing conditions. Mobility shifts were demonstrated in all mutant EspAs as a consequence of slight charge differences between arginine and the residues substituted and likely conformational changes induced by the a position substitutions. Additionally, single and double mutations, with the exception of Arg 163 (pICC25), appeared to affect the multimerization potential of the protein, as evidenced by depletion of some oligomeric forms (Fig. 5).
Mutations in the Coiled-coil Domain Affect EspA Filament Formation-The results of the immunostaining of the EspA filaments is summarized in Table II and illustrated in Fig. 6. This revealed greatly shortened surface protrusions, which appeared as "bobbles" or stumps around the cell in UMD872(pICC25) and UMD872(pICC26) compared with UMD872(pMSD2). In contrast, and as suggested by the altered mobility pattern on native gels, the double position a amino acid mutants lacked any EspA filament structure. The observation that all double substitutions at different positions along the length of the predicted coiled-coil domain in EspA resulted in the same mutant phenotype (Table II) provides a strong indication that the underlying basis for the inability of the EspA mutant to assemble into a filament involves disruption of a structural domain rather than resulting from more serious perturbations in the predicted ␣-helical secondary structure of the protein. The fact that the double mutant EspAs are soluble and secreted further support this suggestion (data not shown).
A/E lesion Formation on Cultured HEp2 Cells-In order to identify possible functional effects of the coiled-coil mutagenesis, we investigated the ability of the different strains to induce A/E lesions on cultured HEp-2 cells using the FAS test as a marker for lesion formation (23). Single amino acid substitution by arginine at position Leu 149 or Met 163 was insufficient to inhibit formation of A/E lesions by UMD872(pICC26) or UMD872(pICC25), respectively, whereas double mutants, even after extended incubations (5 h), were unable to initiate host cytoskeletal rearrangements and A/E lesion formation (Table  II; Fig. 6).
Conservative Reversion of Arginine Mutations-To demonstrate that the mutant phenotypes were solely a consequence of the specific substitutions introduced at the carboxyl terminus, we targeted the non-permissive Arg 149 residue in pICC26 for substitution with more conservative amino acids methionine, phenylalanine, and valine producing plasmids pICC30, pICC31, and pICC32, respectively. A similar methionine substitution was also introduced into the double mutant pICC27 at the same position generating plasmid pICC33.
The FAS phenotype and EspA filament assembly were restored in UMD872(pICC31) and UMD872(pICC32) (Table II; Fig. 6), while UMD872(pICC30) remained deficient of wild type   (Table II). All the revertant strains showed a positive FAS. Restoration of EspA multimerization in these revertant strains was similarly reflected by the characteristic wild type "laddering" profile of EspA protein in native gels (Fig.  5, lanes 6 and 7). The ability to promote A/E lesions was similarly restored in UMD872(pICC33) but filament structure in this strain was indistinguishable from the vestigial structures encoded by pICC25 and pICC26 complements. Taken together these results provide a strong indication of functional restoration of a structural domain in the EspA protein that is proposed to be important for interactions leading to filament assembly.

DISCUSSION
Coiled-coils form important structural domains in diverse proteins mediating oligomerization of monomers via association of two or more ␣-helices. Sequences that are capable of forming coiled-coils are characterized by heptad repeats comprising seven amino acids designated a-g, where the first (a) and fourth (d) positions are occupied by hydrophobic amino acids, and the remaining positions (b, c, e, f, and g) by polar amino acids. The a and d position residues form a hydrophobic interface between helices which are buried in the core of the assembled coiled-coil complex. The hydrophobic interactions form the basis for the stability of the complex.
The prevalence of predicted coiled-coils among proteins of type III secretion systems (16) hints at a potential mechanism by which proteins may interact in the formation of multimeric complexes or processes involved in the translocation of effector proteins. The prediction of a coiled-coil segment in a region of EspA that demonstrates significant similarity to a well characterized coiled-coil region in flagellins from a number of bacterial species introduces the possibility that coiled-coil interactions may contribute to EspA filament assembly.
In addition to flagellins, the assembly of a number of eukaryotic proteins into filamentous structures is based upon interactions between dimeric coiled-coil domains (15). However, only one coiled-coil domain is predicted in the carboxyl terminus of EspA, which presents the possibility that this region mediates homo-and/or hetero-oligomeric interactions between different polypeptides (16). Our results suggest that EspA: EspA interaction is impaired by mutation at an essential carboxyl-terminal domain, although some oligomeric forms of EspA are still evident in the more extreme double mutants (Fig. 5). This is perhaps not unexpected since other molecular interactions in addition to the potential contribution of coiledcoils are likely to be involved in EspA association. Using GST-EspA and His-EspA, we demonstrated EspA-EspA interaction on solid phase. Unfortunately, due to low expression level of the double mutant GST-EspA fusion, we were unable to determine the effect of the mutations on EspA-EspA interaction using this binding assay.
Carboxyl-terminal mutants are, however, unable to produce surface EspA filaments. This is a direct effect of mutation within the predicted coiled-coil domain and may be a consequence of the destabilizing effect incurred by coiled-coil disruption on homo-oligomeric interactions, or alternatively, abrogation of EspA interactions with additional protein(s) more integral to filament assembly or structure. Mutations affecting flagellar presentation and shape are similarly located in a coiled-coil region of the flagellin (17,26).
Unlike double substitution mutants where EspA:EspA interaction and EspA filament assembly was totally abrogated, single substitution mutants produced a vestigial EspA filament. In this situation, alterations within an essential domain may not be sufficient to prevent EspA:EspA interaction but appear to be sufficient to prevent polymerization into long filaments. Interestingly, the vestigial filaments produced by the single substitution mutants are reminiscent of the filament stumps produced by an espD Ϫ strain UMD870 (14). We have provided evidence that neither EspD nor EspB is required for EspA multimerization, although it is clear that EspD is integral to assembly and or elaboration of the filament on the bacterial cell surface (14). It has recently been determined that EspD is transported to the host cell membrane where it remains exposed and solvent accessible (27). EspD has two predicted coiled regions, spanning residues 147-174, and 336 -368 (16), which could potentially mediate structural interactions with EspA and/or other proteins, presenting a potential mechanism by which the EspA filament could be anchored at the host cell surface.
In the light of this recent information, EspA filament assembly may require the contribution of EPEC proteins, which perhaps reflect further the function of flagellar biogenesis components, such as the flagellar capping protein FliD, which is essential for polymerization of flagellin monomers at the tip of the growing flagellum (28). Similarly, a second gene in the fliD operon of Salmonella typhimurium, fliS, encodes a protein that appears to facilitate the export of flagellin through the growing filament channel (29). Mutation at fliS affects the filament elongation step, resulting in short flagella with a concomitant decrease in the amount of excreted flagellin. Both these effects are mirrored, with respect to EspA, by mutation of the EPEC espD gene (14). Alternatively, aberrant surface presentation of the filament may be due to a failure of a more mechanical , and UMD872pICC31 (E) produced mature EspA filaments and exhibited A/E activity, strain UMD872pICC25 (D) produced only vestigial EspA filaments but still exhibited A/E activity, and strains UMD872 (B) and UMD872pICC27 (F) did not produce EspA filaments and were negative for A/E activity. Magnification: columns 1, ϫ2,000; columns 2 and 3, ϫ1,300. nature involving proteins that anchor the filament at the bacterial cell surface.
Demonstration of the structural integrity of the mutant EspA proteins is an important consideration if inferences are to be made concerning the effect of mutations in an uncharacterized domain. The ability to restore wild type EspA function by reversion of a functionally non-permissive mutation to a more conservative residue (Arg 149 to Val), but one that is still different from wild type (Leu 149 ), provides an indication that the altered phenotypic effects were the consequence of aberrant structural interactions, since low tolerance for substitution at a positions comprising the hydrophobic core in coiled-coils is well documented (30). The smaller Val and Phe residues are less likely to grossly deform a coiled-coil in the way that the larger Met residue at the same position might. It is also possible that Met interacts differently with residues from neighboring peptides.
The results presented herein provide the first analysis of EspA filament assembly assessed in the context of culture supernatants, which in the prototypic strain E2348/69 are known to contain several proteins secreted by the type III secretion system (31). Future studies using purified wild type and mutant EspA proteins selected on the basis of these results will facilitate the further characterization of EspA homo-and hetero-oligomeric interactions and enable structural integrity of mutants to be rigorously assessed by the application of procedures such as circular dichroism spectrophotometry.