Hierarchical Delivery of an Essential Host Colonization Factor in Enteropathogenic Escherichia coli*

Many significant bacterial pathogens use a type III secretion system to inject effector proteins into host cells to disrupt specific cellular functions, enabling disease progression. The injection of these effectors into host cells is often dependent on dedicated chaperones within the bacterial cell. In this report, we demonstrate that the enteropathogenic Escherichia coli (EPEC) chaperone CesT interacts with a variety of known and putative type III effector proteins. Using pull-down and secretion assays, a degenerate CesT binding domain was identified within multiple type III effectors. Domain exchange experiments between selected type III effector proteins revealed a modular nature for the CesT binding domain, as demonstrated by secretion, chaperone binding, and infection assays. The CesT-interacting type III effector Tir, which is crucial for in vivo intestinal colonization, had to be expressed and secreted for efficient secretion of other type III effectors. In contrast, the absence of other CesT-interacting type III effectors did not abrogate effector secretion, indicating an unexpected hierarchy with respect to Tir for type III effector delivery. Coordinating the expression of other type III effectors with cesT in the absence of tir partially restored total type III effector secretion, thereby implicating CesT in secretion events. Collectively, the results suggest a coordinated mechanism involving both Tir and CesT for type III effector injection into host cells.

Many significant bacterial pathogens use a type III secretion system to inject effector proteins into host cells to disrupt specific cellular functions, enabling disease progression. The injection of these effectors into host cells is often dependent on dedicated chaperones within the bacterial cell. In this report, we demonstrate that the enteropathogenic Escherichia coli (EPEC) chaperone CesT interacts with a variety of known and putative type III effector proteins. Using pull-down and secretion assays, a degenerate CesT binding domain was identified within multiple type III effectors. Domain exchange experiments between selected type III effector proteins revealed a modular nature for the CesT binding domain, as demonstrated by secretion, chaperone binding, and infection assays. The CesT-interacting type III effector Tir, which is crucial for in vivo intestinal colonization, had to be expressed and secreted for efficient secretion of other type III effectors. In contrast, the absence of other CesTinteracting type III effectors did not abrogate effector secretion, indicating an unexpected hierarchy with respect to Tir for type III effector delivery. Coordinating the expression of other type III effectors with cesT in the absence of tir partially restored total type III effector secretion, thereby implicating CesT in secretion events. Collectively, the results suggest a coordinated mechanism involving both Tir and CesT for type III effector injection into host cells.
Bacterial pathogens often have multiple systems to disrupt and subvert host cellular processes that are involved in disease production. One such system is the type III secretion system (T3SS) 4 that is widespread among Gram-negative pathogens of animals and plants. T3SSs are composed of multiprotein complexes in the bacterial membrane that mediate the rapid injection of type III effector proteins directly into host cells (1).
Enteropathogenic Escherichia coli (EPEC) is a human diarrheal attaching and effacing (A/E) pathogen that attaches to intestinal microvilli and then injects type III effector proteins directly into host cells (2,3). After bacterial attachment, dramatic cytoskeletal rearrangements result in the effacement of microvilli (4). The A/E phenotype has been linked to the locus of enterocyte effacement (LEE) (5), a pathogenicity island that is involved in EPEC virulence (6). The LEE encodes components of the T3SS, transcriptional regulators, chaperones and type III effector proteins, the latter of which are translocated directly into host cells (7).
EPEC and other related strains have multiple type III effectors that are encoded within the LEE, in addition to non-LEEencoded effectors that are located in distinct pathogenicity islands throughout the chromosome (3,8,9). One of the best studied EPEC type III effectors is Tir (translocated intimin receptor), a protein that is injected into host cells, modified by host kinases, and localized to the host membrane (10,11). Furthermore, the amino-and carboxyl-terminal regions of Tir interact with a number of host proteins, causing dramatic host cytoskeletal rearrangements, which result in actin-rich lesions termed pedestals (12)(13)(14)(15)(16). Remarkably, host membrane-localized Tir also serves as a receptor for EPEC via intimin (10), a bacterial outer membrane adhesin also encoded by the LEE of A/E pathogens. The direct importance of Tir in the virulence of A/E pathogens has been convincingly demonstrated in three in vivo animal models of infection, where tir mutants do not colonize the host intestine or cause clinical symptoms of disease (7,17,18). Other well studied LEE encoded type III effectors include Map, EspF, and EspG (19 -22), which have been shown to have multifunctional disruptive properties within host cells.
The translocation of type III effectors into host cells is often dependent on a dedicated family of proteins termed type III chaperones (NCBI conserved domain data base, pFam05932). These proteins are cytosolic or membrane-associated, generally small (ϳ20 kDa), soluble, and negatively charged (pI 4 -5) (23). Type III secretion chaperones typically form homodimers that bind to the amino-terminal region of effectors and remain in the bacterial cell following translocation of effectors into the host cell. The crystal structures of many type III secretion chaperones have been solved, some in a complex with an effector (24 -29). Collectively, these studies have revealed remarkable chaperone structural similarity, even in the absence of primary sequence sim-ilarity. Certain type III secretion chaperones are thought to bind a single effector, whereas others have been demonstrated to have multivalent properties and bind more than one effector.
In the case of EPEC, the LEE encoded type III secretion chaperone CesT was initially shown to bind and stabilize Tir within the bacterial cell (30,31); however, additional studies have demonstrated interactions with the effectors Map, EspF, and NleA (32)(33)(34). CesT is also required for the efficient in vitro type III secretion of other LEE and non-LEE type III effectors (33), suggesting that other chaperone-effector interactions may occur within the bacterium. Furthermore, we have previously demonstrated that CesT interacts with the membrane-associated ATPase EscN of the T3SS, in a role probably serving to recruit and target type III effectors for translocation into host cells (33,35). Interestingly, cesT mutants do not cause overt disease in a mouse model of attaching and effacing pathogenesis (7), indicating that this type III secretion chaperone has a central role in disease.
In this study, we demonstrate that CesT interacts with at least eight EPEC type III effectors. A degenerate CesT binding domain within multiple type III effectors was identified and experimentally confirmed by pull-down and protein domainexchanging experiments. A variety of complementary genetic and biochemical assays were used to demonstrate a modular nature for the CesT binding domain. Last, using multiple EPEC strains in secretion and infection assays, it is demonstrated that a coordinated Tir-CesT interaction is required for the efficient injection of other type III effectors into host cells. The results highlight events leading to the hierarchical injection of the critical host colonization factor Tir, which probably precedes the delivery of subsequent effectors.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-EPEC 2348/69 (36) and relevant mutant strains were used for all experiments unless otherwise indicated (Table 1). Cultures were routinely grown in Luria broth (10 g of Bacto-peptone, 5 g of yeast extract, 10 g of NaCl) and then subcultured into Dulbecco's modified Eagle's medium (Hyclone). E. coli DH5␣ served as a cloning strain and E. coli BL21(DE3) was used as an overexpression strain for selected proteins.
Recombinant DNA Techniques and Mutant Strain Construction-Full-length nleH was amplified from genomic DNA in a PCR using primers 162 (GGAATTCCATATGT-TATCGCCCTCTTCTATAAATTTGG) and 169 (CGGG-TACCTATCTTACTTAATACTACAC). The PCR fragment was restriction-digested with NdeI and KpnI and cloned into the respective restriction sites of pFLAG-CTC (Sigma) to create an in frame fusion with a sequence encoding the FLAG peptide. The resulting construct (pNT243) encodes NleH-FLAG from the recombinant tac promoter of pFLAG-CTC.
To create a plasmid for nleH expression from its native promoter, primers 163 (TCGAATTACCCTGATTTTGCTGG) and 169 were used in a PCR with genomic DNA. The resulting PCR fragment encompassed 419 bp upstream of the nleH start codon and the complete nleH open reading frame. The fragment was restriction-digested with KpnI followed by directional cloning into NruI/KpnI-treated pNT242 to create pNT244, which expresses NleH-FLAG from the nleH promoter. Truncated forms of NleH-FLAG were created by using pNT244 as template in a PCR with primer 163 individually Briefly, an nleH promoter region was amplified from pNT244 with primers 163 and 170 (GGAATTCCATATGTTTCAAC-CTTCAAAATAAACCTATC). An nleH gene fragment was then amplified with primer 173 (GGAATTCCATATG-TATCGTGTTGTGGTCACTGATAATAAG) and primer 169. The two DNA fragments were digested with NdeI and then mixed in a 1:1 ratio, followed by the addition of T4 ligase. An aliquot of the ligation was added to a PCR with primers 163 and 169 to amplify a 1.2-kb fragment that was then treated with KpnI, followed by cloning into NruI/KpnI-digested pNT242 to generate pNT250. The generation of plasmids expressing CBD exchange derivatives of Tir and NleH used a multiple step blunt ligation-PCR strategy. For the Tir-NleH[CBD]-Tir fusion, three fragments were ligated as follows. The NleH CBD (aa 1-40) was PCRamplified with primers 172 (ATGTTATCGCCCTCTTC-TATAAATTTGG) and 176 (GCTATCAGAGTGAACAG-CAGC) using pNT244 as template, generating fragment A (120 bp). A DNA fragment containing nucleotides 250 -1653 of tir was amplified by PCR from EPEC genomic DNA with primers 177 (TGCTTGCTTGGAGGATTTG) and 151 (CCG-GATATCTTAAACGAAACGTACTGGTCC), generating fragment B (1.4 kb). Fragments A and B were treated with T4 polynucleotide kinase, gel-purified, and then blunt end-ligated to each other in a ligation reaction. The ligation reaction then served as template in a PCR with primers 172 and 151, generating fragment C. The promoter and first 114 nucleotides of tir were then amplified from EPEC genomic DNA with primers 145 (GTAAGGAGACTAAATGTCGC) and 174 (AATTA-GATGACCAGTTCCTC), generating fragment D. Fragments C and D were kinase-treated and blunt end-ligated as before. The ligation reaction then served as a template in a PCR with primers 145 and 151, generating a 2.2-kb DNA fragment that was digested with BamHI and subsequently cloned into BamHI/EcoRV-treated pACYC184 to create pNT251. For the Tir[CBD]-NleH fusion, the Tir CBD was PCR-amplified with primers 178 (GGAATTCCATATGAGCTCTACAGGAG-CATTAGGA) and 179 (TGTCTCAGATGTAGCTGCAGC), generating fragment E. A DNA fragment containing nucleotides 121-909 of nleH was PCR-amplified with primers 180 (GGGACGCAAGTAACGGTTGGC) and 169 using pNT244 as template, generating fragment F. The fragments were kinasetreated and then ligated. The ligation reaction served as template in a PCR with primers 178 and 169, which generated a 1-kb fragment that was directionally cloned into NdeI/KpnIdigested pNT250, creating pNT252.
A plasmid construct expressing NleH-FLAG from the tir promoter was created by amplifying a DNA fragment with primers 175 (CCGCTCGAGGTCTGTTAGGAATAATTA-GATAGG) and 169 in a PCR with pNT244 serving as template, followed by cloning of the product into XhoI/KpnI-digested pNT242, to create pNT253. The same primers were used in a PCR with pNT250 as template to create pNT254 expressing NleH[⌬CBD]-FLAG from the tir promoter. A plasmid construct expressing NleA-FLAG from a tac promoter was generated by amplifying a DNA fragment from EPEC genomic DNA with primers 157 (GGAATTCCCATATGAACATTCAAC-CGATCGTAAC) and 158 (CGGGTACCGACTCTTGTT-TCTTGGATTATATC) followed by cloning into NdeI/KpnIdigested pFLAG-CTC to create pNT255.
To create a plasmid expressing Tir (⌬2-38), an inverse PCR strategy was used. Briefly, primer 186 (CATACATATATC-CTTTTATTTAGAAATTTGACACG) and primer 191 (AGCTGTGCACCGTATCGCGG) were used in a PCR with pTir as template DNA. The resulting product was treated with T4 polynucleotide kinase, followed by ligation and transformation into DH5␣, creating pTir(⌬2-38). The plasmid was then transformed into ⌬tir for phenotypic analyses.
A nonpolar in frame nleA deletion mutant was generated by allelic exchange with a nleA gene fragment with an internal in-frame deletion. Briefly, primers 159 (CGGAGCTCGACG-CACTCGACATCTCACTGG) and 160 (CCGCTCGAGGAT-TCCGGATGTTACGATCGG) were used to PCR-amplify a 1013-bp fragment from EPEC genomic DNA. Similarly, primers 161 and 162 generated a 799-bp fragment. The two DNA fragments were digested with XhoI and ligated with T4 DNA ligase. The ligation served as template in a PCR with primers 159 and 162 with the resulting 1.8-kb product being cloned into SacI/KpnI-digested pRE112, generating pNT256. pNT256 was conjugated into EPEC ⌬sepD, followed by the isolation and genotypic verification of ⌬sepD⌬nleA strains, as previously described (37). Similarly, a pRE112-based espZ deletion construct was created using primer pairs EPespZ(1) (GCGGTACCTGCT-TGTCGAGCAACGAGGCG) and ⌬EPespZ(R) (CCGCTAGCG-GATTAGCGATGAAATATGCC) and ⌬EPespZ(F) (GCGCTA-GCTGGTAATACTGCACCAGAAGG) and EPespZ(2) (CCGA-GCTCGAGTATCTTTGTATATTGACTC), followed by the isolation of ⌬espZ mutants. An EPEC ⌬sepD⌬escN mutant was created in the ⌬escN genetic background by allelic exchange using a sepD deletion construct (38) harboring a gene fragment of sepD with an internal in-frame deletion. ⌬sepD⌬escN mutants were verified by PCR genotyping and evaluating protein secretion profiles.
Transcriptional Analyses-The nleA gene promoter was cloned upstream of the promoterless chloramphenicol acetyltranferase gene within pKK232-8 (Amersham Biosciences). The resulting construct was transformed into various EPEC strains and tested in transcriptional reporter assays as previously described (41).
Evaluation of Total EPEC Secreted Proteins-Total EPEC secreted proteins were collected as previously described (33). Briefly, Dulbecco's modified Eagle's medium-cultured EPEC cells were pelleted, and the culture supernatant was then filtered (0.22 m). The filtrate was precipitated by the addition of ice-cold trichloroacetic acid (final concentration 10% (v/v)) and incubated on ice for 30 min. The precipitated proteins were pelleted by centrifugation at 16,000 ϫ g and then washed with ice-cold acetone. Precipitated proteins were resuspended in 2ϫ ESB (0.0625 M Tris-HCl (pH 6.8), 1% (w/v) SDS, 10% glycerol, 2% 2-mercaptoethanol, 0.001% bromphenol blue) and subjected to SDS-PAGE and Western blotting analyses.
CesT Affinity Column Binding and FLAG Immunoprecipitation Pull-down Assays-A CesT affinity column was generated by purifying overexpressed His-CesT from an E. coli BL21(DE3) lysate using Ni 2ϩ -nitrilotriacetic acid affinity chromatography (33). Purified CesT (200 g) was equilibrated with binding buffer (20 mM Tris (pH 7.9), 0.5 M NaCl), followed by the addition of various filtered EPEC culture supernatants. The column was extensively washed with binding buffer, followed by a higher stringency wash buffer (20 mM Tris (pH 7.9), 0.5 M NaCl, 60 mM imidazole). Retained proteins were co-eluted with His-CesT using elution buffer (20 mM Tris (pH 7.9), 0.5 M NaCl, 1 M imidazole). For mass spectrometry identification of CesT-interacting proteins, the procedure was modified to specifically exclude CesT from the elution fraction. Briefly, after CesT-interacting EPEC proteins were column-bound, the column was treated with denaturing binding buffer (20 mM Tris (pH 7.9), 0.5 M NaCl, 6 M urea). This served to denature and elute proteins from the column. The eluate was then passed over a separate Ni 2ϩ -nitrilotriacetic acid column, preequilibrated with denaturing binding buffer, to specifically retain denatured His-CesT. The eluate from this column was then trichloroacetic acid-precipitated, followed by mass spectrometry analyses (see below).
To evaluate a direct interaction between NleH and CesT, FLAG-tagged NleH expressed in DH5␣ was immunoprecipitated from a cell lysate (5-ml culture, A 600 ϭ 0.8) onto FLAG beads as previously described (33). Purified GST-CesT or GST only was added to the NleH-containing beads, followed by incubation at 4°C. The beads were then washed three times with phosphate-buffered saline containing protease inhibitors (Roche Applied Science complete mini-EDTA, one tablet per 10 ml of phosphate-buffered saline). NleH-FLAG was competitively eluted from the beads using FLAG peptide (final concentration of 90 ng/l). The eluted protein fraction was then subjected to SDS-PAGE and immunoblotting analyses. An additional control for GST-CesT demonstrated that it did not bind to FLAG-beads exposed to a DH5␣ lysate, ruling out CesT interacting with the beads or any nonspecific bead-interacting proteins from DH5␣.
In Vitro Cell Culture EPEC Infection Assays-For immunofluorescence microscopy experiments, HeLa cells were infected with overnight LB-cultured EPEC strains at a multiplicity of infection of 100. All infections were carried out at 37°C, 5% CO 2 for 3 h in Dulbecco's modified Eagle's medium containing fetal bovine serum unless otherwise indicated. Infected HeLa monolayers were washed with phosphate-buffered saline, followed by fixation with paraformaldehyde (2.5%, v/v). The samples were processed for immunofluorescence microscopy as previously described (37). Monoclonal antibodies were used to stain fixed samples: anti-FLAG (1:2000) (Sigma), anti-Tir (1:500), and anti-PY (1:500) clone 4G10 (Upstate Biotechnology, Inc.), followed by extensive washing and staining with secondary antibodies (Alexa 488-or 568-conjugated anti-mouse (Molecular Probes), 1:400). For visualization of polymerized actin, Alexa 488-conjugated phalloidin (Molecular Probes) was included with the secondary antibodies at a 1:100 dilution. Images were detected using a Zeiss Axioskop microscope, captured with an Empix DVC1300 digital camera, and analyzed using Northern Eclipse imaging software.
Mass Spectrometry Analyses-20 ml of a filtered EPEC ⌬sepD secreted protein preparation was applied to a CesT affinity column. Retained proteins were extensively washed and eluted as described above. The eluted proteins were trichloroacetic acidprecipitated and washed in acetone before SDS-PAGE, followed by Sypro Ruby Red staining. Well defined dominant protein bands were excised, whereas less obvious protein-containing regions within the same gel lane were excised according to apparent molecular weight. Excised gel slices were subjected to trypsin digestion and peptide recovery using a Montage in-gel digest kit (Millipore) as directed by the manufacturer. The peptides were analyzed on an API Q STAR PULSARi Hybrid liquid chromatograph/tandem mass spectrometer at the University of British Columbia Michael Smith Laboratories/Laboratory of Molecular Biophysics Proteomics Core Facility. The data were analyzed using Mascot software (39), and peptide sequences were FIGURE 1. A, the identification of CesT-interacting type III effectors using an affinity column approach. Protein species in the ⌬sepD supernatant that were retained on the CesT affinity column are denoted with an asterisk. After extensive washing, proteins were eluted with 6 M urea, subjected to SDS-PAGE, and stained with Sypro Ruby Red. Gel slices (numbered 1-8) were excised and subjected to trypsin digestion and mass spectrometry analyses for identification (see Table 2). M, protein standards. B, total secreted proteins derived from the culture supernatant of a type III secretion-deficient mutant, ⌬sepD⌬escN, did not interact with the CesT affinity column. C, detection of a direct NleH-CesT protein interaction using a FLAG tag immunoprecipitation assay. NleH-FLAG from a DH5␣ E. coli lysate was immunoprecipitated onto FLAG M2-agarose beads followed by the addition of GST-CesT or GST only.
searched within the nonredundant Proteobacteria data base at the NCBI. In addition, matches to peptides were also searched within the assembled EPEC 2348/69 genome data base at the Welcome Trust Sanger Institute.

CesT Interacts with Diverse Type III Effector Proteins Encoded on Different Pathogenicity Islands in EPEC-
CesT is known to interact with four effector proteins, Tir, Map, EspF, and NleA (30 -34). Recently, we demonstrated that the efficient secretion of many other EPEC type III effector proteins also required CesT (32,33), suggesting a possible interaction of these proteins with CesT. To identify new CesT interacting type III effector proteins, we took advantage of a ⌬sepD strain of EPEC that is known to hypersecrete type III effector proteins but not translocator proteins (38). ⌬sepD was cultured under conditions that promote type III effector secretion, and the culture supernatant was collected and passed over a CesT affinity column. After extensive washing, bound protein species were eluted and subjected to mass spectrometry analyses.
Tir was completely depleted from the ⌬sepD supernatant flow-through fraction, indicating a strong interaction with CesT (Fig. 1A). NleA was reduced in the flow-through, as well as many other protein species. A ⌬sepD⌬escN strain defective for type III secretion due to the absence of the T3SS ATPase did not secrete known type III effectors and served as a control on a separate CesT affinity column. A column without CesT did not bind supernatant proteins to any significant extent (data not shown). Notably, EspC, a large type V secreted protein (110 kDa) (40), was not depleted to any significant extent from the ⌬sepD flow-through, further indicating the specificity of the column for CesT interactions (Fig. 1A). As expected, mass spectrometry analyses of gel-extracted proteins from the ⌬sepD supernatant identified three known CesT binding partners, Tir, NleA, and EspF. Five other type III effectors (EspG, EspZ, NleG, NleH, and NleH2) ( Table 2; see "Experimental Procedures") were also identified as CesT-interacting proteins. These proteins are known to be substrates of the type III secretion system, and some require CesT for their efficient secretion in EPEC (33). Thus, their observed interaction with the CesT affinity column is consistent with their dependence on CesT for efficient secretion. The novel effectors are encoded in different pathogenicity islands within EPEC (Table 2), implicating CesT as a multivalent chaperone for diverse type III effector proteins.
Identification of the CesT Binding Domain of NleH-The retention of multiple type III effectors on the CesT affinity column could be mediated by direct CesT-type III effector inter- actions or bridging interactions between proteins. To examine the possibility that CesT affinity column interactions were due to bridging proteins, NleH-FLAG was expressed from a plasmid in E. coli DH5␣ that does not encode EPEC type III effectors. Notably, NleH-FLAG could be expressed as a soluble protein in the absence of CesT, which was not the case for all type III effectors tested (see "Discussion"). NleH-FLAG was immunoprecipitated onto FLAG beads from a cell lysate and then mixed with purified recombinant GST-CesT (30), followed by extensive washing and elution with FLAG peptide. GST-CesT co-eluted with NleH-FLAG, whereas GST alone did not bind to the NleH-FLAG-containing beads (Fig. 1B). Next, we set out to further characterize the CesT interaction with NleH. Plasmidencoded NleH was expressed in ⌬sepD or ⌬sepD⌬cesT and examined for protein secretion. NleH was highly secreted in ⌬sepD but only to a minimal extent in ⌬sepD⌬cesT, as demonstrated by Western blotting total secreted protein preparations, confirming the requirement of CesT for efficient NleH secretion ( Fig. 2A). Carboxyl-terminal truncated versions of NleH and a deletion mutant (⌬2-50 NleH) were then expressed in ⌬sepD to characterize the type III secretion requirements of NleH. All of the ⌬sepD strains harboring plasmids had similar secretion profiles (Fig. 2B). NleH (aa 1-50) was not detectable in lysate or secreted protein samples, whereas the longer versions were all secreted at high levels from ⌬sepD (Fig. 2C). The amino-terminal deletion mutant (⌬2-50 NleH) was not secreted, and the protein remained in the bacterial cell, as demonstrated by Western blotting of ⌬sepD whole cell lysates (Fig.  2D). Next, the culture supernatant from EPEC ⌬sepD expressing NleH (aa 1-100) was collected and passed over the CesT affinity column. NleH (aa 1-100) was depleted from the flowthrough fraction and remained bound after extensive washing. The protein was co-eluted with CesT (Fig. 2E), indicating that the CesT binding domain of this effector resides within the first 100 amino acids.
Collectively, these experiments demonstrate an interaction between CesT and the amino-terminal region of NleH. Furthermore, the first 100 amino acids of NleH are necessary and sufficient to mediate CesT-dependent secretion of this non-LEE-encoded type III effector in EPEC.
A Putative Degenerate CesT Binding Domain Is Located at the Amino Terminus of Diverse Type III Effectors-The ability of column-bound CesT to retain multiple type III effectors suggested that each protein has a functional CBD. In order to determine if this binding domain was conserved between the type III effectors, the known CBD of Tir was used as a "scaffold" to identify similar sequences in other type III effectors. Amino acid sequences of other known and putative EPEC type III effectors were evaluated by alignment or multialignment analyses (e.g. BLAST or Clustal, default settings); however, no significant sequence similarities were observed (data not shown). Altering the ClustalW parameters revealed an alignment with the amino-terminal region of NleH and NleH2 of EPEC (Fig. 3A). The alignment of the amino-terminal regions of other EPEC and EHEC type III effectors (from different pathogenicity islands)   also produced an alignment revealing conserved amino acids, suggesting that this region could represent the CBD for these effectors (Fig. 3B).
The Degenerate CesT Binding Domain Is Modular in Nature and Can Be Functionally Exchanged between Effectors-Based on the CesT-dependent protein secretion of multiple type III effectors and CesT interaction data presented here, it was hypothesized that the CBD is modular in nature and serves to target effectors for type III secretion via interaction with CesT. To test this hypothesis, the CBD of NleH and Tir were swapped (see "Experimental Procedures" for chimeric protein construction). The respective plasmid constructs were transformed into EPEC for characterization studies.
Deletion of the CBD from NleH produced a stable protein species that was not secreted to any extent (Fig. 2B). Fusing the CBD of Tir (amino acids 39 -83) to NleH lacking its own CBD resulted in the recombinant NleH protein being secreted at levels comparable with wild type NleH (Fig. 4A). This indicated that the Tir CBD, when fused to NleH (missing its native CBD), can support the type III secretion of the recombinant protein in EPEC.
Since the CBD of Tir is located ϳ38 amino acids into the protein, the CBD of NleH (aa 1-40) was used to replace amino acids 38 -83 of Tir. Deletion of the CBD of Tir destabilizes the protein, consistent with Tir stability being strictly dependent on the presence of CesT in EPEC (30). The recombinant Tir protein harboring the NleH CBD was rendered stable and was efficiently secreted by ⌬tir EPEC (Fig.  4A), indicating that the CBD of NleH functionally served to stabilize cytoplasmic Tir and supported its secretion. A ⌬escN mutant defective for type III secretion did not secrete the NleH-and Tir-interchanged CBD fusion proteins (data not shown). This latter result indicated that the secretion of the fusion proteins remained dependent on a functional type III secretion system.
The ability of each fusion protein to interact with CesT was examined using the aforementioned CesT affinity binding assay. The Tir-NleH[CBD]-Tir fusion protein secreted from ⌬tir bound to the CesT column at levels similar to native Tir secreted by wild type EPEC (Fig. 4B). The Tir[CBD]-NleH fusion was also observed to interact with CesT. Bacteria expressing the recombinant proteins were then tested in an infection assay for their ability to deliver the domain-exchanged type III effectors into HeLa cells. Tir injection into HeLa cells results in the formation of actin-rich pedestals that can be observed as punctate actin staining by immunofluorescence. Since the Tir-NleH[CBD]-Tir hybrid protein was stable and efficiently secreted in secretion assays, we evaluated the efficiency of pedestal formation for a tir mutant expressing the recombinant Tir-NleH[CBD]-Tir hybrid protein. The ⌬tir mutant expressing a Tir derivative lacking its CBD did not produce pedestals or focus actin to any significant extent, whereas the tir mutant harboring pNT251 encoding Tir-NleH[CBD]-Tir produced pedestals and focused actin similar to wild type EPEC (Fig. 5). Wild type EPEC harboring a plasmid encoding epitope-tagged NleH or the Tir[CBD]-NleH fusion translocated the respective NleH proteins into HeLa A B C FIGURE 6. The presence of Tir is important for efficient type III effector secretion. A, total secreted protein profiles of various EPEC strains. Note that the deletion of tir from the ⌬sepD background dramatically reduces type III effector secretion. The ⌬sepD⌬tir double mutant was transcomplemented with a plasmid expressing SepD to return the total secreted protein profile back to that of ⌬tir. Transcomplementation of ⌬sepD⌬tir with plasmid-expressed Tir partially restored type III effector secretion to ⌬sepD levels. The arrowheads point to protein species in the last lane, which are lacking in the ⌬sepD⌬tir secretion profile. M, protein standards. B, deletion of other effector genes encoded within the LEE (espZ) or outside the LEE (nleA) does not alter total type III effector secretion. The arrowheads indicate protein species that are not present in the ⌬sepD⌬tir secretion profile. C, transcriptional fusion analyses using a plasmid nleA reporter construct harbored within different EPEC strains. cells (Fig. 5). NleH and Tir[CBD]-NleH both localized immediately underneath adherent bacteria, close to regions of considerable actin focusing. These experiments indicate that the degenerate CesT binding domain is modular in nature and can be functionally exchanged between type III effectors. Tir Deletion Abrogates Type III Effector Secretion but Not Gene Expression-The tir and cesT genes are co-transcribed as part of a multicistronic transcript from the LEE5 promoter (41). The co-transcription and adjacent arrangement of many effector-chaperone pairings are common features among pathogens, and it is believed that the coordinated expression perhaps allows for the productive interaction of the respective proteins. It is now known that not all effector genes are co-transcribed with their partner type III secretion chaperone, since some are located within different pathogenicity islands within the chromosome. We asked the question whether CesT could carry out its function for other effectors in the absence of Tir protein expression. The ⌬sepD strain hypersecretes effectors that are readily detectable by Coomassie staining and thus represents an ideal genetic background to address the role of Tir with respect to other type III effectors. Therefore, a ⌬sepD⌬tir strain was generated and evaluated for effector secretion. Surprisingly, total effector secretion was dramatically reduced for the ⌬sepD⌬tir strain compared with the ⌬sepD parent (Fig. 6A). Notably, abundant levels of NleA were missing from the culture supernatant as well as other type III effector proteins. The ⌬sepD⌬tir strain was trans-complemented with a plasmid expressing Tir, which restored total effector secretion. This experiment also served to rule out polar transcriptional effects on chromosomal cesT gene expression. Deletion of either a type III effector gene encoded outside of the LEE (nleA) or within the LEE (espZ) from the ⌬sepD background did not alter secreted type III effector levels (Fig. 6B), indicating that Tir protein expression is important to sustain total type III effector secretion. A transcriptional fusion to nleA was used to evaluate the expression level of this effector gene in different EPEC strain backgrounds. nleA gene expression was not significantly different in ⌬sepD or ⌬sepD⌬tir strains (Fig. 6C), indicating that the absence of NleA secretion in ⌬sepD⌬tir was not due to a lack of nleA transcription.
A Coordinated CesT-Type III Effector Interaction Rescues the ⌬sepD⌬tir Secretion Defect-Based on these results, it was hypothesized that by unlinking tir from cesT gene expression, the CesT protein would be temporally unbound and that free CesT would act as a feedback signal to negatively modulate the secretion of other type III effectors. It was therefore hypothesized that any type III effector that interacts with CesT should return total type III effector secretion to the ⌬sepD⌬tir strain, provided that its expression was linked with CesT. To address this hypothesis, NleA-FLAG was expressed within ⌬sepD⌬tir from a recombinant tac promoter, which is active during typical Tir expression conditions. NleA-FLAG (encoded by pNT255) was found to be secreted and partially restored the secretion of other type III effectors (Fig. 7). Another plasmid encoding NleH from the tir promoter (pNT253) was transformed into the ⌬sepD⌬tir strain and evaluated for effector secretion. NleH expressed from the tir promoter also partially restored type III effector secretion minus Tir (Fig. 7). Notably, NleA secretion was restored. Last, a plasmid encoding NleH without its CBD (pNT254) did not restore effector secretion, suggesting that a coordinated CesT-type III effector protein interaction is important for efficient type III effector secretion.
Tir Secretion Is Required to Efficiently Activate Type III Effector Secretion in EPEC-The ability of any effector to rescue total effector secretion in the ⌬sepD⌬tir double mutant could be due to an interaction with CesT. Alternatively, other type III effectors may require the action of Tir secretion from the cytoplasm to support their subsequent secretion. To address these questions, we constructed a plasmid that expresses a CesTinteracting Tir variant without its putative signal sequence (Tir ⌬2-38 aa). This region of Tir encodes the putative Tir secretion signal (aa 1-26) (42) up to the beginning of the Tir CBD. Immunoblotting of cell lysates and the secreted protein preparations revealed that the Tir variant was stable within the EPEC cytoplasm, suggestive of an interaction with CesT. In addition, this Tir variant was only minimally secreted compared with wild type Tir (Fig. 8A). Notably, this Tir variant did not rescue effector secretion, whereas overexpression of a wild type Tir protein FIGURE 7. The coordinated expression of NleA or NleH with CesT (achieved by using the recombinant tac or tir promoter, respectively) restored type III effector secretion and required an effector-CesT interaction. The arrowheads indicate protein species that are restored to the ⌬sepD⌬tir secretion profile by transcomplementation with pNT253 and pNT255 but not pNT250. Note that the protein bands in the pNT253 and pNT255 transcomplemented ⌬sepD⌬tir secretion profiles are not Tir breakdown products. The presence of similar molecular weight protein species present in the same amount in the ⌬sepD⌬nleA double mutant (Fig. 6) indicates that these proteins species are not NleA breakdown products. These protein species were identified by mass spectrometry and are listed in Table  2. M, protein standards.
completely restored type III effector secretion as visualized by staining total secreted proteins.
The Putative Tir Secretion Signal Cannot Impose Type III Effector Secretion Hierarchy-Interestingly, amino acids 38 -75 of Tir aligns with amino acids 2-40 of NleH (Fig. 3A). This region of both proteins corresponds to the CBD, as determined by the domain exchange experiments presented here. The absence of a similar secretion signal within NleH (and other type III effectors) combined with the observed hierarchy of Tir secretion over other type III effectors led us to hypothesize that amino acids 1-38 of Tir (which overlap with the putative signal sequence) (42) mediate Tir secretion hierarchy. Therefore, a plasmid construct that expresses Tir (aa 1-38) fused to NleH transcribed from the nleH promoter was created and tested in secretion assays. ⌬sepD was found to secrete comparable levels of NleH and Tir (aa 1-38)-NleH as determined by immunoblotting of total secreted protein preparations (Fig. 8B). In addition, Tir (aa 1-38)-NleH did not restore total type III effector secretion to ⌬sepD⌬tir (Fig.  8C), indicating that amino acids 1-38 of Tir alone cannot impose secretion hierarchy to other type III effectors.

DISCUSSION
Type III secretion chaperones bind effector proteins, promoting their rapid and efficient translocation into host cells. Here, it is demonstrated that a degenerate protein domain found in multiple EPEC type III effectors is involved in CesT binding. Interestingly, the data demonstrate that the coordinated interaction of the type III effector Tir with its chaperone CesT, followed by Tir secretion, appears to be a sequence of events that is absolutely critical for the efficient secretion of other type III effectors. This finding is in agreement with Tir being a critical in vivo colonization factor and for the first time elucidates a hierarchy for Tir delivery over other EPEC type III effectors.
A pressing question in bacterial pathogenesis is how pathogens regulate their multiple virulence determinants to create a productive infection leading to disease. For extracellular intestinal pathogens, it is absolutely critical to attach and colonize gut tissues before being prematurely shed into the environment. We show here with genetic and biochemical evidence that the receptor for intimate attachment, Tir, is the hierarchical type III effector in EPEC. The data suggest that once Tir-mediated intimate attachment has been achieved, then other type III effectors involved in subverting the host's intracellular processes are injected. The results are in direct accordance with animal infection experiments where Tir-mediated adherence of A/E pathogens is important for host colonization and disease progression. This point has been directly and robustly demonstrated in three animal models of infection: (i) EHEC tir mutants do not efficiently colonize the infant rabbit intestine and do not show clinical signs of disease (18); (ii) Citrobacter rodentium tir mutants do not colonize the mouse intestine and do not cause disease (43); and (iii) rabbit FIGURE 8. Tir secretion is required to efficiently activate type III effector secretion in EPEC. A, total secreted protein preparations from various EPEC strains were subjected to SDS-PAGE, followed by Coomassie Blue staining. The lower panel shows an anti-Tir immunoblot using one-tenth of the sample amount loaded for Coomassie Blue staining. Note that Tir (⌬2-38) is minimally secreted by the ⌬sepD⌬tir strain and does not return the secretion profile back to ⌬sepD levels. M, protein standards. B, fusion of the putative Tir secretion signal (aa 1-38) to the amino terminus of NleH does not increase secretion levels of the recombinant NleH protein, as demonstrated by SDS-PAGE of total secreted proteins (top) and immunoblotting (bottom). C, NleH or a Tir (aa 1-38)-NleH fusion expressed from the nleH promoter did not rescue total type III effector secretion within ⌬sepD⌬tir, as demonstrated by SDS-PAGE of total secreted proteins followed by Coomassie Blue staining. OCTOBER 5, 2007 • VOLUME 282 • NUMBER 40

JOURNAL OF BIOLOGICAL CHEMISTRY 29643
EPEC tir mutants do not cause diarrhea in weaned rabbits and do not intimately adhere to host intestinal cells (17).
Previously, a comprehensive screen of LEE-interacting proteins using the yeast two-hybrid assay identified a strong CesT-Map and a weak CesT-EspF interaction (34) but failed to identify EspG and EspZ as interacting proteins. Interestingly, EspF is reported to have its own chaperone CesF (44) (rorf10 within the LEE). We cannot rule out the possibility that EspF interacts with both CesF and CesT. Our results independently confirm a CesT-EspF interaction using a different approach ( Table 2). We also identify a putative CesT binding domain on EspF through comparative analyses (Fig. 3).
There could be additional CesT-interacting type III effectors that were not secreted at high enough levels within our EPEC secretion assays or were in an improper conformation to interact with immobilized His-CesT (Fig. 1). Indeed, it has been reported that the CesT-interacting type III effector Map is expressed at very low levels within EPEC (32), which may explain why it was not detected in our analyses. Nonetheless, from our observations and those from other reports (32-34), CesT physically interacts or is involved in the secretion of every LEE-encoded effector described to date, further strengthening its central role in EPEC pathogenesis. Pairwise protein interaction studies between CesT and each of its partner effectors will help to identify specific binding affinities for this multivalent chaperone of EPEC.
The multivalent binding property of CesT with different type III effectors suggests that there is a common feature among the effectors that mediates CesT binding. A study by Lilic et al. (26) suggests that a number of type III effectors have a conserved structural ␤-fold that type III secretion chaperones interact with. This does not appear to be the case for all of the EPEC type III effectors that interact with CesT, since the presumed structural ␤-folds are not in a proper spatial context for dimeric CesT binding. From our CBD swapping experiments with EPEC effectors, it is clear that a degenerate binding domain of ϳ40 amino acids is involved in CesT binding.
We hypothesize that EPEC has a hierarchical mechanism to deliver its receptor initially as modeled in Fig. 9. The coordinated expression of tir and cesT genes promotes a timely protein interaction of the CesT chaperone dimer with Tir within the bacterial cell, which results in efficient Tir secretion. In the absence of Tir (e.g. tir mutants), CesT is probably uncomplexed and perhaps is able to have a negative feedback mechanism with respect to effector secretion. We tested this possibility by overexpressing CesT in both wild type and tir mutant strains during secretion permissive growth conditions and observed no difference compared with strains with normal CesT levels (data not shown). This suggests that CesT alone does not mediate regulated type III effector secretion. We hypothesize that the action of Tir secretion results in CesT being associated with the secretory apparatus. This is in agreement with the observed inner membrane association of CesT within EPEC, a localization that is enhanced in the presence of Tir (33). After Tir secretion, other effectors can bind to membrane-associated or cytoplasmic CesT, which then supports their secretion through the secretory apparatus (Fig. 9). This model is similar to a proposed model for Yersinia pestis type III effector secretion, where it is thought that LcrQ and its chaperone SycH function together to mediate a hierarchy of secretion (45).
Prior to this study, genetic evidence for type III effector hierarchy was lacking, although convincing evidence for the hierarchical assembly of the type III translocon has been reported (46). From the studies presented here, it is evident that certain type III effectors are hierarchical over others. The data presented here also implicate type III secretion chaperones to be involved with hierarchy, although additional studies are required to elucidate the mechanisms and dynamics underlying this process. The role of multivalent secretion chaperones, such as CesT, are particularly interesting in this regard, given their multiple interactions with diverse type III effectors.