Enterohaemorrhagic Escherichia coli (EHEC) O157:H7 uses a specialized protein translocation apparatus, the type III secretion system (TTSS), to deliver bacterial effector proteins into host cells. These effectors interfere with host cytoskeletal pathways and signalling cascades to facilitate bacterial survival and replication and promote disease. The genes encoding the TTSS and all known type III secreted effectors in EHEC are localized in a single pathogenicity island on the bacterial chromosome known as the locus for enterocyte effacement (LEE). In this study, we performed a proteomic analysis of proteins secreted by the LEE-encoded TTSS of EHEC. In addition to known LEE-encoded type III secreted proteins, such as EspA, EspB and Tir, a novel protein, NleA (non-LEE-encoded effector A), was identified. NleA is encoded in a prophage-associated pathogenicity island within the EHEC genome, distinct from the LEE. The LEE-encoded TTSS directs translocation of NleA into host cells, where it localizes to the Golgi apparatus. In a panel of strains examined by Southern blot and database analyses, nleA was found to be present in all other LEE-containing pathogens examined, including enteropathogenic E. coli and Citrobacter rodentium, and was absent from non-pathogenic strains of E. coli and non-LEE-containing pathogens. NleA was determined to play a key role in virulence of C. rodentium in a mouse infection model.
Enterohaemorrhagic Escherichia coli (EHEC) O157:H7 is a recently emerged pathogen that causes diarrhoea, which can progress to haemorrhagic colitis and haemolytic uraemic syndrome (Nataro and Kaper, 1998). EHEC belongs to a group of Gram-negative pathogens defined by their ability to form attaching/effacing (A/E) lesions on host cells during infection (Frankel et al., 1998; Nataro and Kaper, 1998). A/E pathogens include enteropathogenic E. coli (EPEC), a diarrhoeal pathogen especially prevalent in developing countries, and several animal pathogens including Citrobacter rodentium and rabbit enteropathogenic E. coli (REPEC) (Vallance et al., 2002a). Attachment of A/E pathogens to intestinal epithelial cells induces disassembly of microvilli in infected cells and the formation of actin-rich pedestals underneath adherent extracellular bacteria (Frankel et al., 1998). The genes required for the formation of A/E lesions are clustered together in a single chromosomal pathogenicity island known as the locus for enterocyte effacement (LEE), which is present in all A/E pathogens (Elliott et al., 1998; Perna et al., 1998; Deng et al., 2001; Zhu et al., 2001; Tauschek et al., 2002). The LEE contains genes that encode a type III secretion system (TTSS) and several translocated effector proteins.
TTSSs are multisubunit molecular machines that are present exclusively in Gram-negative pathogens and symbionts but absent from commensal bacteria (Hueck, 1998). The TTSS has been described as a ‘molecular syringe’: it spans both inner and outer bacterial membranes and, upon contact with host cells, assembles a translocon that forms a continuous channel for the delivery of effector proteins directly into host cells. The genes encoding type III apparatus components are conserved between bacterial species, but the translocated effector proteins tend to be unique to each pathogen (Hueck, 1998). TTSS-translocated effectors modulate host cell pathways to promote bacterial survival and replication within the host, and the set of effectors translocated by each pathogen reflects the unique needs and specific niches of each bacterial species.
A/E pathogens use the LEE-encoded TTSS to insert their own receptor for intimate attachment into host cells (Kenny et al., 1997). The translocated intimin receptor (Tir) is translocated into host cells where it forms a hairpin structure in the host cell plasma membrane with two transmembrane domains, an extracellular loop and amino- and carboxy-terminal domains that are exposed to the host cell cytoplasm (Kenny et al., 1997; de Grado et al., 1999). The extracellular loop of Tir interacts directly with the bacterial outer membrane protein intimin, thus anchoring the bacteria tightly to the host cell (de Grado et al., 1999). The cytoplasmically exposed domains of Tir bind host cytoskeletal and signalling proteins and initiate actin polymerization at the site of bacterial attachment (Goosney et al., 2001; Gruenheid et al., 2001). This results in the formation of actin pedestal structures underneath adherent bacteria.
Four other LEE-encoded TTSS-translocated effectors have been characterized in A/E pathogens. EspH enhances the elongation of actin pedestals (Tu et al., 2003); EspF plays a role in disassembly of tight junctions between intestinal epithelial cells (McNamara et al., 2001); EspG is related to the Shigella microtubule-binding effector VirA (Elliott et al., 2001; Yoshida et al., 2002); and Map localizes to mitochondria (Kenny and Jepson, 2000), but also has a role in actin dynamics (Kenny et al., 2002). A myriad of other TTSS-dependent effects of A/E pathogens on host cells have been described in the literature; however, the effectors responsible for many of these phenotypes have not been identified (Goosney et al., 1999; Celli et al., 2001; de Grado et al., 2001; Nougayrede et al., 2001; Dahan et al., 2002; Shifrin et al., 2002). The number of TTSS-translocated effector proteins identified in A/E pathogens remains small (five) compared with other pathogens such as Salmonella, and their roles do not account for all the observed aspects of the disease caused by these organisms. To gain further insight into other type III-mediated effects on host cells that are distinct from pedestal formation, we used proteomics to identify the first non-LEE-encoded effector translocated by the LEE-encoded TTSS.
Identification of NleA
Although type III secretion is generally thought to be contact dependent (Hueck, 1998), defined in vitro culture conditions can induce EHEC to secrete type III effectors into extracellular medium during growth in liquid culture (DeVinney et al., 1999; Li et al., 2000). Culture supernatants were prepared from wild-type EHEC (wt) and a type III secretion mutant (escN–), grown in type III secretion-inducing conditions. Analysis of the secreted proteins by SDS-PAGE revealed one abundant high-molecular-weight protein common to the secreted proteins from both wild-type and escN– samples, and several other abundant proteins unique to the wild-type sample (Fig. 1A). The secreted proteins were separated by two-dimensional gel electrophoresis, and the abundant separated protein spots were excised from the gel and analysed by mass spectrometry (Fig. 1B, Table 1). Spot ♯1, which was present in both wild-type and escN– culture supernatants, was identified as EspP, a plasmid-encoded protein of EHEC that is secreted by an autotransporter mechanism independently of type III secretion (Brunder et al., 1997). Four major spots (♯2, 3, 4 and 5) were unique to the wild-type supernatants. Three of these spots were identified as known type III secreted proteins encoded within the LEE: Tir (spot ♯2), EspB (spot ♯4) and EspA (spot ♯5) (Fig. 1B, Table 1). Spot ♯3 was identified as a protein of predicted molecular weight of 48 kDa encoded by an open reading frame (ORF) within the EHEC genome but outside the LEE (Table 1, Fig. 2A). Owing to subsequent experiments (see below), we called this protein NleA, for non-LEE-encoded effector A.
Table 1. . Mass spectrometric results for secreted protein spots.
No. of peptides
Predicted MW (kDa)
Experimental MW (kDa)
Characterization of the locus containing nleA
Examination of the locus containing the nleA gene showed that it is encoded in an O-island, a region of the EHEC genome absent from the genome of the non-pathogenic E. coli strain K-12 (Perna et al., 2001). This island encompasses the insertion site of a prophage, CP-933P, and has many characteristics associated with pathogenicity islands (PAIs) containing horizontally transferred genes (Hacker and Kaper, 2000). The region between the last gene conserved in the E. coli K-12 backbone (yciE) and genes encoding phage structural proteins contains several putative transposase gene fragments and one putative site-specific recombinase gene fragment (Fig. 2A). Analysis of this region with islandpath, a program designed to identify PAIs (Hsiao et al., 2003), revealed that all ORFs within this region have a dinucleotide bias and a GC content divergent from the EHEC genome mean (Fig. 2A). Together, these results strongly indicate that nleA is localized to a PAI containing horizontally transferred genes. Several other ORFs within this region have interesting features suggesting roles in virulence. Z6021 encodes a protein that is 27% identical over a region of 148 amino acids to OspG, a protein of unknown function that was identified on the basis of its secretion by the Mxi–Spa TTSS of Shigella flexneri (Buchrieser et al., 2000). Further upstream, the region encompassing Z2565–Z6012 is homologous to a previously described operon of EPEC that is co-transcriptionally regulated with the LEE region via the plasmid-encoded transciptional activator Per/BfpT (Tobe et al., 1999). The first ORF of this operon, trcA, is proposed to encode a cytoplasmic chaperone for intimin and BfpA, the major subunit of the bundle-forming pilus of EPEC. However, EHEC does not encode a bundle-forming pilus nor does it encode Per/bfpT, suggesting that the regulation and/or function of this operon may be distinctive in EHEC. Analysis of the in-progress genome sequence (http://www.sanger.ac.uk/Projects/Microbes/) indicates that, in EPEC, the trcA-containing operon does not map near nleA or near the LEE. However, in the rabbit enteropathogenic E. coli strain 84/110-1, trcA is immediately adjacent to the LEE. These data suggest that genomic rearrangement of pathogenicity islands has occurred in A/E pathogens and/or that some of these islands may be mosaic in origin.
To investigate further the nature and distribution of the nleA gene, an EHEC nleA probe was prepared, and Southern blots were performed on a panel of genomic DNA from other A/E pathogens and a non-pathogenic E. coli strain. As shown in Fig. 2B, the nleA gene was present in all A/E pathogens examined, but absent from the non-pathogenic E. coli. strain. Analysis of the in-progress EPEC genome sequence revealed that nleA is present in close proximity to a phage insertion site in the EPEC genome. Additional database searches revealed that nleA is present within a prophage of an intimin-positive, non-O157 EHEC strain, O84:H4, but absent from non-pathogenic strains of E. coli and uropathogenic E. coli, which do not contain the LEE. nleA is also absent from other TTSS-containing pathogens such as Salmonella and Shigella species. These data suggest that nleA has been specifically acquired or retained in A/E pathogens. We used polymerase chain reaction (PCR) to amplify the nleA gene from Citrobacter rodentium (submitted to GenBank, accession ♯AY430401) and acquired the EPEC and O84:H4 nleA gene sequences from the databases. A multiple sequence alignment of the four predicted protein sequences reveals a high degree of sequence conservation in these four A/E pathogens (Fig. 2C).
NleA is secreted by the LEE-encoded type III secretion system
All EHEC and EPEC proteins secreted by the LEE-encoded TTSS described to date are also encoded within the LEE. However, it has been demonstrated that other Gram-negative pathogens that use a TTSS, such as Salmonella Typhimurium, possess type III-secreted effectors that are encoded at a distance from their cognate secretion systems (Galan, 2001). To confirm that secretion of NleA was dependent on the LEE-encoded TTSS, an epitope-tagged version of NleA was expressed from a plasmid in wild-type EHEC and an escN– strain, which is deficient for type III secretion (Jarvis and Kaper, 1996). As shown in Fig. 3A, although haemagglutinin (HA)-tagged NleA was expressed to similar levels in wild-type and escN– EHEC, the protein was only secreted into the extracellular media by the pnleA-HA-transformed wild-type bacteria. DnaK, an abundant non-secreted bacterial protein, was used as a control for the absence of non-secreted proteins in the secreted protein samples (Fig. 3B). Tir was secreted in the untransformed and pnleA-HA-transformed wild-type strains, but not by the escN– strains (Fig. 3C), verifying the expected TTSS phenotypes. Similar results were obtained for the expression of epitope-tagged NleA in wild-type EPEC and several type III secretion mutants of EPEC, indicating that NleA can also be secreted by the EPEC TTSS (data not shown).
NleA is translocated into host cells
When EHEC is grown under type III secretion-inducing conditions, two types of proteins are secreted into the extracellular medium. In addition to translocated effectors such as Tir, there are also translocators: components that facilitate the translocation of effector proteins into host cells (DeVinney et al., 1999; Li et al., 2000). These include EspA, which has been shown to form a filament between bacteria and host cells through which proteins are translocated (Knutton et al., 1998; Daniell et al., 2001; Neves et al., 2003), and EspB and D, which participate in the formation of a translocation pore in the host cell membrane (Ide et al., 2001). To distinguish whether NleA was a translocator or a translocated effector, we investigated type III secretion and translocation in the absence of NleA by generating a deletion mutant strain. Secreted protein profiles from wild-type and a ΔnleA mutant EHEC strains were analysed. The wild-type sample contained an abundant protein of ≈ 50 kDa that was absent from the ΔnleA secreted proteins (Fig. 4A). Western blot analysis with antisera directed against NleA demonstrated that 50 kDa NleA was present in the wild-type secreted proteins and absent in the ΔnleA sample (Fig. 4B). However, other than the presence or absence of NleA, the secreted protein profiles of the wild-type and ΔnleA strains were identical (Fig. 4A). Thus, NleA is not required for secretion of other type III-secreted proteins.
To determine whether NleA is required for translocation of other type III effectors into host cells, HeLa cells were infected with wild-type EHEC and EHEC ΔnleA, and Tir translocation and function were monitored by immunofluorescent staining of infected cells. As shown in Fig. 4C, EHEC ΔnleA adhered to HeLa cells at similar levels to wild-type EHEC (Fig. 4C, far left). Immunofluorescent staining revealed that Tir was translocated into host cells and focused under infecting bacteria in both wild-type and ΔnleA EHEC strains (Fig. 4C, middle left). To confirm functional Tir translocation, infected cells were stained with fluorescent phalloidin to visualize polymerized actin involved in pedestal formation underneath adherent bacteria (Fig. 4C, middle right). Actin pedestals were evident in cells infected with either wild-type or ΔnleA EHEC, indicating that translocation and function of other type III effectors can proceed in the absence of NleA. These experiments also indicate that NleA is not required for pedestal formation.
As NleA did not appear to play a role in the secretion or translocation of other effectors, we investigated whether NleA was translocated itself. HeLa cells were infected for 6 h with wild-type or escN– EHEC expressing HA-tagged NleA and subjected to subcellular fractionation and Western blot analysis with an anti-HA antibody. As indicated in Fig. 5A, NleA is translocated into host cells where it associates with the host cell membrane fraction. Translocation of NleA is not observed during infection of cells with a type III secretion mutant expressing HA-tagged NleA, indicating that NleA translocation and host membrane association is TTSS dependent. Western blotting of the fractions with antibodies to proteins specific to each fraction confirmed the absence of cross-contamination of the fractions. Calnexin, a host cell integral membrane protein, was absent from the host cytoplasmic fraction, and tubulin, a host cell cytoplasmic protein, was absent from the host membrane fraction. DnaK, a non-secreted bacterial protein, was present only in the low-speed pellet, demonstrating a lack of bacterial contamination of the host membrane and cytosolic fractions. NleA and DnaK were absent from the low-speed pellet in the type III mutant-infected cells as a result of the type III dependence of EHEC adherence. As EPEC adherence to HeLa cells is independent of type III secretion, we performed further control experiments expressing and delivering NleA-HA by wild-type, Δtir, ΔespA and type III secretion-deficient EPEC. As expected, NleA was translocated to the membrane fraction of cells infected with wild-type and Δtir strains, but not ΔespA or type III mutant strains (data not shown). Thus, NleA translocation into host cells requires a functional TTSS (including a functional translocon), but does not require Tir.
To investigate the nature of NleA association with host cell membranes, infected host cell membrane fractions containing HA-tagged NleA were extracted on ice under several conditions and recentrifuged to obtain soluble and insoluble membrane fractions (Fig. 5B). These fractions were subjected to Western blot analysis with anti-HA antibody to detect HA-tagged NleA. Treatment with high salt (1 M NaCl) or alkaline pH (0.2 M Na2CO3, pH 11.4) removes proteins that are peripherally associated with membranes via electrostatic or hydrophilic interactions respectively. The association of NleA with host cell membranes resisted disruption with these treatments (Fig. 5B, top), as did calnexin, an integral membrane protein (Fig. 5B, middle). In contrast, a significant proportion of calreticulin, a peripheral membrane protein, was extracted from the membrane fraction during both high-salt and alkaline pH treatment (Fig. 5B, bottom). Treatment of membrane fractions with the non-ionic detergent Triton X-100, which solubilizes integral membrane proteins such as calnexin (Fig. 5B, middle), almost completely solubilized NleA, resulting in a shift of the HA-tagged NleA protein from the insoluble to the soluble fraction (Fig. 5B, top). These results indicate that NleA is translocated into host cells where it behaves as an integral membrane protein. Indeed, analysis of the NleA protein sequence by several transmembrane domain prediction programs predicts one or two putative transmembrane domains within the sequence (Fig. 2C).
NleA localizes to the host Golgi apparatus
The subcellular localization of NleA within host cells was then determined. As shown in Fig. 6A, immunofluorescent staining of HeLa cells infected with wild-type EHEC, using the anti-NleA antibody, resulted in a perinuclear pattern of staining that was absent in cells infected with EHECΔnleA (Fig. 6D) or uninfected cells (data not shown). This pattern did not resemble staining obtained with markers for late endosomes, lysosomes, endoplasmic reticulum (ER), mitochondria or the nucleus (data not shown). However, a very similar pattern of staining was observed when cells were co-stained with anti-NleA (Fig. 6A) and antibodies to markers of the Golgi apparatus, including mannosidase II (Fig. 6B), where the two proteins co-localized extensively (Fig. 6C). To confirm Golgi localization of NleA, infected cells were incubated with brefeldin A, a fungal metabolite that disrupts Golgi structure (Chardin and McCormick, 1999), before fixation and immunofluorescence. As shown in Fig. 6G and H, brefeldin A treatment caused diffusion of both mannosidase II and NleA staining, as expected for Golgi-localized proteins. Co-localization of NleA was observed with several other markers of the Golgi apparatus, and Golgi localization was also observed in experiments examining epitope-tagged NleA stained with anti-tag antibodies, using both HA and FLAG epitope tags (data not shown). To determine whether Golgi localization of NleA required other bacterial factors or was an inherent property of NleA, cells were transfected with an expression construct encoding a green fluorescent protein (GFP)–NleA fusion protein. As shown in Fig. 6, transfected NleA–GFP (Fig. 6J) also localized to the Golgi, where it overlapped with mannosidase II staining (Fig. 6K and L).
NleA is required for virulence
The high degree of sequence conservation of nleA in EHEC and C. rodentium (Fig. 2C) suggests that NleA plays a similar role in EHEC and C. rodentium infections. C. rodentium is a natural pathogen of mice (Schauer and Falkow, 1993) and has been used as a model system to study A/E pathogenesis. The C. rodentium model has been used previously to demonstrate key roles for the TTSS (Vallance et al., 2002b) and type III-secreted effectors such as Tir (Deng et al., 2003) in virulence. In susceptible strains of mice, C. rodentium infection is fatal, and typically causes death of infected mice between days 6–10 of infection (Vallance et al., 2003). More resistant mouse strains do not die from C. rodentium infection, but become colonized and develop intestinal inflammation and colonic hyperplasia (Vallance et al., 2003).
To test the role of NleA in virulence, we created an nleA-deleted C. rodentium strain and verified the absence of NleA by Western blotting total bacterial extracts with NleA antiserum (Fig. 7A). Mice were infected with equal numbers of wild-type or ΔnleA bacteria by oral gavage. In C. rodentium-susceptible C3H-HeJ mice, NleA was absolutely required for virulence. All C3H-HeJ mice infected with wild-type C. rodentium died between day 6 and 10 after infection (n = 9), whereas all ΔnleA-infected mice (n = 13) displayed some mild disease symptoms such as soft stools but still gained weight and were active throughout the infection and survived indefinitely (Fig. 7B). Notably, C3H/HeJ mice infected with the ΔnleA strain were resistant to subsequent challenge with wild-type C. rodentium when challenged at 6 weeks or 10 weeks after the initial infection (n = 5 at 6 weeks, n = 5 at 10 weeks; 100% survival; data not shown). Thus, although the ΔnleA strain is severely attenuated in susceptible mice, it interacts sufficiently with the host to stimulate protective immunity.
To ensure that the attenuation of ΔnleA C. rodentium was not caused by an unlinked mutation or a polar effect of our mutation on other genes, the nleA mutation was complemented with nleA under the control of its own promoter, on a medium/low-copy-number plasmid. Western blotting of total bacterial extracts with NleA antiserum verified restored expression of NleA in the complemented strain (Fig. 7A). When C3H/HeJ mice were infected with the complemented strain, the lethality of infection was restored, and all animals in the complemented group died between day 9 and 13 after infection (n = 7, Fig. 7B). The slight delay in the time of death of the mice infected with the complemented strain compared with the wild-type strain may result from the effects of expression of nleA from a multicopy plasmid rather than from a single genomic copy of the gene. However, despite the small delay in time of death of mice in the complemented group, expression of nleA from a plasmid clearly complemented the defective virulence of the ΔnleA bacteria, proving that the effect of the nleA mutation on virulence is specific.
In contrast to C3H/HeJ mice, C. rodentium infection is not lethal for outbred NIH Swiss mice. In these mice, C. rodentium colonization of the large intestine leads to intestinal inflammation, colonic hyperplasia and mild diarrhoeal symptoms. NIH Swiss mice were infected with wild-type C. rodentium or the ΔnleA strain and sacrificed at day 10 after infection. The mice infected with the ΔnleA strain had, on average, a 20-fold lower C. rodentium titre in the colon at day 10 (mean values ± SE: 9 ± 2 × 107 cfu for wild type; 4 ± 2 × 106 cfu for ΔnleA; t-test P-value 0.01; Wilcoxon test P-value 0.001; Fig. 7C). In histological analyses of biopsies taken from the anal verge of infected mice, numerous bacteria were evident in the wild type-infected tissue (Fig. 7G), but bacteria were scarce in the ΔnleA-infected samples (Fig. 7H, white arrow). All animals infected with wild-type C. rodentium displayed pathological signs of colonic hyperplasia, whereas all ΔnleA-infected mice had no signs of hyperplasia and displayed relatively normal histology (Fig. 7E and F). The relative degree of intestinal inflammation and hyperplasia was also evident in the difference in colon weights at the time of sacrifice in the two groups of mice (mean values ± SE: 290 ± 13 mg for wild type; 183 ± 7 mg for ΔnleA; t-test P-value 0.00003; Fig. 7D). The wild type-infected mice also had significantly larger spleens than the ΔnleA-infected mice (mean values ± SE: 174 ± 9 mg for wild type; 128 ± 7 mg for ΔnleA; t-test P-value 0.003; Fig. 7D).
To characterize further the virulence defect of ΔnleA bacteria, we performed mixed infection experiments to determine the competitive index of ΔnleA during C. rodentium infection of mice. The competitive index is a sensitive in vivo measure of the contribution of a given gene to virulence that assays colonization ratios of wild-type and mutant bacteria within the same animal (reviewed by Beuzon and Holden, 2001). To determine the competitive index, inoculum containing approximately equal numbers of wild-type and ΔnleA bacteria were mixed and used to inoculate both C3H/HeJ and NIH Swiss mice. The exact input ratio of mutant to wild-type bacteria was later determined by serial dilution and plating of the input inoculum. At day 4 after infection (C3H/HeJ), and day 10 after infection (NIH Swiss), mice were sacrificed, and the ratio of mutant to wild-type bacteria in infected mouse colons was determined by serial dilution and plating followed by genotyping individual bacterial colonies at nleA, using PCR with primers flanking the deleted region of the ΔnleA mutant. The competitive index was then determined as the output ratio (mutant:wild type) divided by the input ratio (mutant:wild type). In C3H/HeJ mice at day 4, the competitive index of ΔnleA was determined to be 0.02. In NIH Swiss mice at day 10, no ΔnleA bacteria were observed out of the 50 colonies examined per mouse, indicating that the competitive index of ΔnleA under these conditions is < 0.02. Therefore, the ΔnleA strain is severely attenuated, and its defect cannot be complemented by co-infection with wild-type bacteria.
It has been problematic to reconcile the number of type III-dependent effects of A/E pathogens on host cells with the apparent paucity of type III-secreted effectors described to date. Five translocated effectors have been described within the LEE, and systematic genetic analyses of all ORFs within the LEE indicate that no other LEE-encoded proteins are translocated effectors (Tu et al., 2003; W. Deng and B. B. Finlay, unpublished). Additionally, although it is well understood how the LEE mediates intimate attachment of bacteria to host cells, the cause of disease during infection with A/E pathogens is still unclear. Using a proteomic rather than a genetic approach to study type III-secreted proteins, we have identified the first non-LEE-encoded effector of A/E pathogens secreted by the LEE-encoded TTSS. Whereas other Gram-negative pathogens that use TTSSs such as Salmonella spp. possess type III-secreted effectors that are encoded apart from their cognate secretion systems (Galan, 2001), this has not yet been described for A/E pathogens such as EPEC and EHEC. Our discovery of a non-LEE-encoded LEE TTSS substrate raises the possibility that other type III-translocated effectors may be encoded elsewhere in the EHEC genome. Thus, the number of proteins translocated by the LEE TTSS may be underestimated. A large number of substrates has been described for other TTSSs, such as the Salmonella SPI-1 TTSS, which secretes at least 19 different proteins (Galan, 2001). We are currently investigating less abundant secreted proteins present in our preparations in order to identify other TTSS substrates. Our identification of NleA and perhaps other TTSS substrates opens up new areas of investigation to increase our understanding of EHEC- and EPEC-mediated diseases.
It is interesting to note that a 54 kDa protein with N-terminal sequence identical to NleA was previously reported to be secreted in sepL mutant but not wild-type EHEC (Kresse et al., 2000). At that time, the sequence obtained for this protein was limited, and the genome sequence for EHEC was not yet available, confounding further identification or characterization of the protein. The reason why the protein was not observed to be secreted from wild-type EHEC by Kresse et al. (2000) is not known, but may reflect technical differences in the experimental procedures, such as bacterial growth conditions.
The presence of the nleA locus in other LEE-containing organisms and the secretion of NleA by the LEE-encoded TTSS provide an interesting example of a functional interaction between separate PAIs within a single pathogen. This has significant implications for the evolution and role of PAIs in A/E pathogens. Phylogenetic analyses suggest that the gain and loss of mobile virulence elements have occurred frequently in pathogenic E. coli (Reid et al., 2000). The origin and acquisition of the LEE in various A/E pathogens are still under investigation. The comparative arrangement of LEE genes and the genomic context of the LEE in various A/E pathogens suggest transfer via a plasmid (Deng et al., 2001). On the other hand, the LEE in EHEC O157:H7 contains prophage sequences (Perna et al., 1998), although these are thought to have been inserted after the mobilization of the LEE into the chromosome (Hacker and Kaper, 2000). It has been proposed recently that the LEE may have originated as a much larger PAI that later underwent recombination events resulting in the LEE and a second PAI in EHEC, known as O-island 122 (Morabito et al., 2003). Indeed, in some A/E pathogens, O-island 122 and the LEE are physically linked (Tauschek et al., 2002; Morabito et al., 2003). It is formally possible that the LEE and nleA-containing island were also acquired simultaneously and later underwent recombination, although we do not yet have any data to support this. It is also possible that nleA was acquired independently of the LEE but somehow adapted to use the pre-existing TTSS for translocation into host cells. Such functional integration of nleA with other pre-existing virulence mechanisms would provide a means for the pathogen to benefit directly from the newly acquired genetic information. Further studies on the presence and genomic context of nleA in other A/E pathogens may clarify the relationship between nleA and the LEE.
Our results indicate that NleA localizes to the Golgi. The observation that a transfected NleA–GFP fusion protein localizes to the Golgi suggests that the NleA protein contains Golgi-targeting information and does not require other bacterial factors to get to this destination. Bacterially delivered NleA is also Golgi localized. However, the NleA protein sequence contains no classical Golgi-targeting motifs (Gleeson, 1998). It is therefore possible that NleA contains novel motifs that direct it to the Golgi, using previously described or novel host cell pathways. Alternatively, NleA could be targeted to the Golgi through a specific interaction with a host cell protein that is itself Golgi targeted. Future studies to determine the mechanism of NleA trafficking in host cells are under way. Other TTS effectors of A/E pathogens localize to the mitochondria (Kenny and Jepson, 2000) and plasma membrane (Kenny et al., 1997), suggesting that the LEE-encoded TTSS is not only involved in intimate attachment of bacteria at the plasma membrane of host cells. Rather, it appears to deploy a multifaceted strike on several host organelles simultaneously. Although the reasons for Golgi localization of NleA are unknown, Golgi localization may provide clues to NleA function. The Golgi is the site for post-translational modification of proteins, such as glycosylation, and also has a central role in protein trafficking, including plasma membrane protein traffic, protein secretion and the sorting of lysosomal proteins. NleA could be targeted to the Golgi to benefit from or interfere with any of these processes. It is possible that NleA targets the Golgi in order to become post-translationally modified or to gain access to other compartments through the Golgi. However, we have been unable to detect modification of NleA in host cells or NleA localization to post-Golgi compartments. An intriguing hypothesis is that NleA affects Golgi function and interferes with host protein modification or trafficking to create a more favourable niche or to defuse host defence mechanisms. We are currently investigating Golgi function during EHEC infection to understand better the role of NleA during pathogenesis.
We have demonstrated a striking effect of NleA on virulence in a mouse model of disease. In susceptible mice, the presence of functional NleA in C. rodentium leads to a lethal infection within 10 days. Mice infected with a strain lacking NleA exhibit few symptoms and survive the infection indefinitely. In a more resistant mouse strain in which C. rodentium infection is non-lethal, NleA is required for the development of colonic hyperplasia and, at day 10 after infection, there are less nleA mutant bacteria present in the host intestine. Although these studies indicate a clear effect of NleA in C. rodentium virulence, the functional role of NleA in infection is still unknown. The number of C. rodentium in the intestine at day 10 represents the outcome of the complex interplay between adherence and replication in the intestine by the bacteria and clearance of the bacteria from the intestine by host resistance mechanisms. We cannot determine from our data where in this balance NleA acts. However, our results from EHEC infection of HeLa cells demonstrate that, in vitro, NleA does not affect adherence of bacteria to host cells or translocation of other effectors, suggesting that NleA may act at the level of resisting host clearance rather than enhancing bacterial adherence. However, the results from our mixed infection experiments demonstrate that attenuation of the nleA mutant bacteria cannot be complemented in trans by co-infection with wild-type bacteria containing functional NleA, arguing against a global effect of NleA on the host intestinal environment. Furthermore, the resistance of ΔnleA-infected mice to subsequent challenge with wild-type C. rodentium provides evidence that an nleA mutant strain colonizes and interacts with the host sufficiently to stimulate host immunity. This is in contrast to type III mutants of C. rodentium that do not colonize the host and provide no protection from subsequent challenge (W. Deng and B. B. Finlay, unpublished). Future studies on the exact role of NleA during infection should provide insight into novel aspects of A/E pathogenesis.
To make EHEC-pnleA-HA, the coding region of nleA was amplified using the proofreading Elongase amplification system (Invitrogen) and the following primers: Z6024F, 5′-AGATCTGAAGGAGATATTATGAACATTCA A C C GACCATAC; Z6024R, 5′-CTCGAGGACTCTTGTTTCTTCGATTATATCAA AG. PCR products were cloned using the TOPO TA cloning kit (Invitrogen), and the DNA sequence was verified using the Taq dye terminator method and an automated 373A DNA sequencer (Applied Biosystems). The product was then subcloned into pCRespG-2HA/BglII, a pACYC-derived plasmid engineered to drive protein expression from a C. rodentium espG promoter and to add two influenza HA to the C-terminus of the expressed protein. The plasmid constructs were then introduced into wild-type EHEC and EHEC escN– by electroporation.
A deletion mutant in nleA in a nalidixic acid-resistant strain of EHEC was created by sacB gene-based allelic exchange (Donnenberg and Kaper, 1991). Two DNA fragments that flank nleA were PCR amplified using EHEC chromosomal DNA as template. Fragment A was PCR amplified using primer NT10 5′-CCGGTACCTCTAACCATTGACGCACTCG and primer NT11 5′-AACCTGCAGAACTAGGTATCTCTAAT GCC to generate a 1.3 kb product. Fragment B was amplified using primer NT12 5′-AACCTGCAGCTGACTATCCTCGT ATATGG and primer NT13 5′-CCGAGCTCAGGTAATGA GACTGTCAGC to generate a 1 kb product. Fragments A and B were then digested with PstI for 1 h, and the enzyme was heat inactivated for 20 min at 65°C. Approximately 50 ng of each digested fragment was added to a ligation reaction with T4 DNA ligase for 1 h at room temperature. The ligation reaction was diluted 1:10, and 1 µl was added to a PCR using primers NT10 and NT13. The resulting 2.3 kb PCR product was then digested with SacI and KpnI, ligated to the corresponding sites of pRE112(Edwards et al., 1998) and transformed into DH5αλpir to generate pNT225. pNT225 was transformed into the conjugative strain SM10λpir, which served as the donor strain in a conjugation with wild-type EHEC. Nalidixic acid- and chloramphenicol-resistant exoconjugants were selected on LB agar. The exoconjugants were then plated onto LB agar containing 5% (w/v) sucrose and no NaCl. The resulting colonies were screened for sensitivity to chloramphenicol, followed by PCR to identify isolates with the nleA deletion and loss of plasmid sequences.
A C. rodentium nleA mutant was generated using a similar procedure. The nleA coding region was amplified from C. rodentium genomic DNA using the EHEC-specific primers Z6024F and Z6024R. C. rodentium nleA flanking DNA sequences were obtained by inverse PCR. Genomic DNA was digested with BglII, followed by ligating the DNA fragments (40 ng ml−1 final concentration) to promote intramolecular ligation. The ligation was diluted 1:10 and used in a PCR with primers NT14 5′-GTATATCCTCTGGAATATGC and NT15 5′-GACATACATCCAACAACCGG. This reaction produced a 0.4 kb product corresponding to sequence upstream and at the 5′ end of nleA. For sequence downstream of nleA, a HindIII genomic DNA fragment was similarly self-ligated and amplified by PCR using primers NT16 5′-GCCATCTGGAG TAAGAGG and NT17 5′-TGCGGTTCACTGACTACC, resulting in a 1.7 kb product. DNA sequence was verified using the Taq dye terminator method and an automated 373A DNA sequencer (Applied Biosystems).
PCR was used to create, in pRE118 (Edwards et al., 1998), a suicide vector bearing an internal deletion of nleA. The following primers were used: del1F 5′-GGTACCACCA CACAGAATAATC; del1R 5′-CGCTAGCCTATATACTGCTGT TGGTT; del2F 5′-GCTAGCTGACAGGCAACTCTTGGACT GG; del2R 5′-GAGCTCAACATAATTTGATGGATTATGAT. The resulting plasmid was introduced into C. rodentium by electroporation to create an antibiotic-resistant merodiploid strain. Loss of plasmid sequences through a second recombination event was selected for as described above. Antibiotic-sensitive, sucrose-resistant colonies were verified for the proper recombination event by PCR using primers flanking the deleted region (CitroHKpro 5′-CGGAATTCCCACAACCAAT CCAAATAACGC and delR2). The absence of NleA was verified by Western blotting whole-cell lysates with polyclonal anti-NleA antiserum.
To make C. rodentiumΔnleA + pCrNleA, the C. rodentium nleA coding region and 5′ flanking region were amplified from C. rodentium genomic DNA using PCR and the primers CitroHKpro and CitroR2 5′-CCCGGGTTAGACTCTTGT TTCTTGGATTATTTCAAAG. The product was cloned using the TOPO TA cloning kit (Invitrogen), and the DNA sequence was verified using the Taq dye terminator method and an automated 373A DNA sequencer (Applied Biosystems). The product was then subcloned into pACYC184, and the plasmid was introduced into C. rodentiumΔnleA by electroporation. Restoration of NleA expression was verified by Western blotting total bacterial extracts using anti-NleA antiserum.
Secreted proteins were prepared as described previously (Li et al., 2000). Wild-type EHEC and EHEC escN– were grown overnight in LB medium. Cultures were then diluted 1:100 into M-9 minimal medium supplemented with 44 mM NaHCO3, 8 mM MgSO4, 0.4% glucose and 0.1% casamino acids and grown standing at 37°C in 5% CO2 to an OD600 of 0.6–0.8. Secreted proteins were harvested by centrifuging cultures at 8000 g for 30 min, thus separating the supernatants from the pellets. Supernatants were filtered through 0.45 micron filters, and the protein concentration was determined by BCA assay (Sigma).
Proteins were prepared for electrophoresis by precipitation with 1/9th volume 100% cold TCA on ice for 45–120 min, followed by centrifugation at 17 600 g for 30 min. The pellets were rinsed in cold 100% acetone and solubilized in 1× Laemmli buffer (for one-dimensional SDS–PAGE gels) or two-dimensional sample buffer for two-dimensional gels (8 M urea, 2 M thiourea, 4% CHAPS, 20 mM Tris, 0.002% bromophenol blue). For two-dimensional gels, dithiothreitol (DTT) was added to 6 mg ml−1 and IPG buffer (pH 3–10; Amersham Biosciences) to 0.5% before loading. Immobiline dry strips (18 cm, pH 3–10; Amersham) were rehydrated in the sample overnight at 20°C. Samples were then focused at 15°C for 65 000 Vh. After focusing, strips were equili-brated in EB + 10 mg ml−1 DTT for 15 min, and then EB + 25 mg ml−1 iodoacetamide for 15 min (EB is 50 mM Tris, 6 M urea, 30% glycerol, 2% SDS, pH 8.8). Equilibrated strips were sealed onto the top of large-format SDS–PAGE gels (12% or 14% acrylamide) using 0.5% agarose in SDS–PAGE running buffer + 0.002% bromophenol blue, and the gels were run until the dye front ran off the gel. Gels were stained with Sypro Ruby according to the manufacturer's instructions (Bio-Rad) and visualized on a UV lightbox. Spots of interest were excised manually.
In-gel digestion of proteins was performed on the Investigator ProGest robot (Genomic Solutions) according to the method of Shevchenko et al. (1996). Samples were analysed by LC-MS/MS on a LCQ Deca ion trap mass spectrometer (Thermo Finnigan) equipped with a Nanoflow liquid chromatography system (LC Packings–Dionex). Reversed-phase PicoFrit columns PFC7515-PP18-5 (New Objective) were used for peptide separation, and the column effluent was sprayed directly into the mass spectrometer. A flow rate of 200 nl min−1 was used, and the total acquisition time was equal to 45 min per sample. Spectra were searched against the NCBI (Bethesda, MD) database with mascot (Matrixscience) or sonar (Proteometrics Canada) search engines.
Southern blot analysis
Genomic DNA samples for Southern blot analysis were prepared using DNeasy tissue kit (Qiagen). Probe was prepared by digesting pNleA-HA with SalI and BglII enzymes to obtain an ≈ 500 bp fragment, which was labelled using a BrightStar Psoralen-Biotin non-isotopic labelling kit (Ambion). Five micrograms of each genomic DNA sample was fully digested with 25 units of BamHI, EcoRI and PstI overnight. The samples were resolved by electrophoresis on 1% agarose gel and transferred overnight to BrightStar-Plus nylon membrane (Ambion) by passive, slightly alkaline downward elution. The DNA was cross-linked to the membrane by exposing the membrane to UV light for 2 min, followed by 30 min of baking at 80°C. The membrane was prehybridized by washing it in 10 ml of ULTRAhyb ultrasensitive hybridization buffer (Ambion) at 42°C for 30 min. Ten microlitres of the prepared probe was then added to the prehybridized membrane in buffer, and the probe was hybridized to the membrane overnight at 42°C. Membrane was washed twice for 5 min in low-stringency wash buffer (Ambion) and twice for 15 min in high-stringency wash buffer (Ambion) at room temperature. The hybridized probe was detected using a BrightStar BioDetect non-isotopic detection kit (Ambion), followed by exposure to Kodak film.
Generation of anti-NleA antiserum
The coding portion of nleA was amplified from EHEC genomic DNA and cloned into a his-tagged expression vector (pET28a; Novagen) using the following primers: 5′-TTCC ATATGAACATTCAACCGACC and 5′-GGAATTCAATAATA GCTGCCATCC. This plasmid was introduced into BL21 (λDE3), grown to an optical density (A600) of 0.8 and induced with 0.5 mM IPTG for 16 h at 20°C. His-tagged protein was purified on a Ni-NTA column according to the manufacturer's instructions (Qiagen). The NleA-containing fractions were pooled, thrombin was added (500:1), and the protein was dialysed overnight against 20 mM Tris, pH 8.2, 50 mM NaCl. The next day, the protein was loaded on a monoQ FPLC column, and the column was developed with a linear gradient from 50 to 500 mM NaCl. NleA-containing fractions were pooled. The protein was > 90% pure after this step. Purified protein was used to immunize two male Sprague–Dawley rats, 300 µg of protein per rat using Freund's complete adjuvant (Sigma), and the resulting antisera was affinity purified using the activated immunoaffinity support Affi-Gel 15 according to the manufacturer's instructions (Bio-Rad). For immunofluorescence experiments, antiserum was further purified by absorption against acetone powders prepared from HeLa cells and from EHEC ΔnleA as described by Harlow and Lane (1988). Specificity of antiserum was confirmed by Western blotting of cell extracts from wild-type EHEC and EHEC ΔnleA.
Samples for Western blot analysis were resolved by SDS-PAGE (9% to 12% polyacrylamide). Proteins were transferred to nitrocellulose, and immunoblots were blocked in 5% non-fat dried milk (NFDM) in TBS, pH 7.2, containing 0.1% Tween 20 (TBST) overnight at 4°C and then incubated with primary antibody in 1% NFDM TBST for 1 h at room temperature. Membranes were washed six times in TBST and then incubated with a 1:5000 dilution of horseradish peroxidase-conjugated goat anti-mouse (H + l) antibody (Jackson ImmunoResearch Laboratories) for 1 h at room temperature. Membranes were washed as described above. Antigen–antibody complexes were visualized with an enhanced chemiluminescence detection kit (Amersham), followed by exposure to Kodak film (Perkin-Elmer). The following primary antibodies were used: anti-HA.11 (Covance), anti-DnaK (Stressgen), anti-EHEC Tir, anti-NleA (this study), anti-calnexin (Stressgen), anti-calreticulin (Affinity Bioreagents) and anti-tubulin (Sigma).
HeLa cells were grown on glass coverslips in 24-well tissue culture plates and infected for 6 h with 1 µl (EHEC) of a standing overnight culture of OD ≈ 0.4. At 6 h after infection, cells were washed three times in PBS containing Ca2+ and Mg2+ and fixed in 2.5% paraformaldehyde in PBS for 15 min at room temperature. Cells were permeabilized in 0.1% saponin in PBS, blocked in 5% goat serum or 5% BSA in PBS + 0.1% saponin and incubated with the following primary antibodies diluted in blocking solution for 1 h at room temperature: anti-EHEC Tir, 1:1000; anti-E. coli O157 (Difco), 1:200; affinity-purified rat polyclonal anti-NleA (this study), 1:100; antimannosidase II (kindly provided by Dr M. Farquhar, UCSD), 1:1000. After three washes in PBS/saponin, cells were incubated in secondary antibodies [Alexa 488- or 568-conjugated anti-mouse, -rabbit, -rat (Molecular Probes) 1:400] for 30 min at room temperature, washed three times in PBS/saponin and once in PBS and mounted onto glass slides using Mowiol + DABCO. For visualization of polymerized actin, Alexa 488-conjugated phalloidin (Molecular Probes) was included with the secondary antibodies at a 1:100 dilution. Where indicated, cells were incubated with 5 µg ml−1 brefeldin A (Boehringer Mannheim) for 30 min before fixation. Images were detected using a Zeiss Axioskop microscope, captured with an Empix DVC1300 digital camera and analysed using northern eclipse imaging software or on a Bio-Rad Radiance Plus confocal microscope using lasersharp software.
Fractionation of infected host cells
For each sample, two confluent 100 mm dishes of HeLa cells were infected with wild-type EHEC-pNleA-HA or EHECescN– pNleA-HA using an initial multiplicity of infection (MOI) of 1:10. At 6 h after infection, cells were washed three times with ice-cold PBS and subjected to biochemical fractionation as described previously (Gauthier et al., 2000; Knodler et al., 2003). Briefly, cells were resuspended in 300 µl of homogenization buffer (3 mM imidazole, 250 mM sucrose, 0.5 mM EDTA, pH 7.4) supplemented with COMPLETE protease inhibitor cocktail (Roche) and mechanically disrupted by passage through a 22-gauge needle. The homogenate was centrifuged at low speed (3000 g) for 15 min at 4°C to pellet unbroken cells, bacteria, nuclei and cytoskeletal components (low-speed pellet). The supernatant was subject to high-speed ultracentrifugation (41 000 g) for 20 min at 4°C in a TLS55 rotor in a TL100 centrifuge (Beckman) to separate host cell membranes (pellet) from cytoplasm (supernatant). The pellets were resuspended in 300 µl of 1× Laemmli buffer, and the supernatant was made up to 1× Laemmli buffer using a 5× stock. Equal volumes of all fractions were resolved by SDS-PAGE (9% polyacrylamide), transferred to nitrocellulose and assayed by Western blot.
For extraction studies of membrane-associated NleA, two 100 mm dishes of infected HeLa cells were fractionated as described above for each extraction condition. The high-speed pellets (host membrane fraction) were resuspended in 300 µl of one of the following extraction buffers: (i) 10 mM Tris, 5 mM MgCl2, pH 7.4; (ii) 10 mM Tris, 5 mM MgCl2, 1 M NaCl, pH 7.4; (iii) 0.2 M NaHCO3, 5 mM MgCl2, pH 11.4; (iv) 10 mM Tris, 5 mM MgCl2, 1% Triton X-100, pH 7.4. Extraction was performed on ice by pipetting the samples up and down every 5 min for 30 min, and the samples were recentrifuged at 100 000 g for 30 min. The pellet (insoluble fraction) was resuspended in 300 µl of 1× Laemmli buffer, the supernatant (soluble fraction) was precipitated in 10% trichloroacetic acid on ice for 30 min, washed in 100% acetone and resuspended in 300 µl of 1× Laemmli buffer. Equal volumes were resolved by SDS-PAGE (9% polyacrylamide), transferred to nitrocellulose and assayed by Western blot.
Infection analysis of C. rodentium in mice
Five-week-old C3H/HeJ mice (Jackson Laboratory) and outbred NIH Swiss mice (Harlan Sprague–Dawley) were housed in the animal facility at the University of British Columbia in direct accordance with guidelines drafted by the University of British Columbia's Animal Care Committee and the Canadian Council on the Use of Laboratory Animals. Wild-type C. rodentium, the nleA deletion mutant and the complemented strain (ΔnleA+ pCrNleA) were grown in LB broth overnight in a shaker at 200 r.p.m., and 100 µl of the cultures was used to infect mice by oral gavage. Inoculum was titred by serial dilution and plating and was calculated to be ≈ 4 × 108 cfu per mouse for all groups. For infection of the highly susceptible C3H/HeJ mice by C. rodentium, the survival of infected mice were assessed daily over the course of the infection. When any mouse became moribund, it was immediately sacrificed. For bacterial virulence assays using NIH Swiss mice, animals were sacrificed at day 10 after infection. To score colonic hyperplasia, the first 4 cm of the distal colon starting from the anal verge was collected and weighed after any faecal pellets were removed. To assay bacterial colonization, colonic tissues plus faecal pellets were homogenized in PBS using a Polytron tissue homogenizer and serially diluted before being plated on MacConkey agar (Difco Laboratories). Colonic tissue and faecal pellets were combined to determine the total bacterial burden in the mouse colon at the time of sacrifice. MacConkey agar is selective for Gram-negative bacteria, on which C. rodentium forms colonies with a highly distinctive and identifiable morphology not typical of E. coli. For histological analysis, the last 0.5 cm of the colon of infected mice was fixed in 10% neutral-buffered formalin, processed, cut into 3 µm sections and stained with haematoxylin and eosin. Histological analysis was done by the Morphological Services Laboratory at the Department of Pathology and Laboratory Medicine of the University of British Columbia. For competitive index experiments, equal volumes of the inoculum of wild-type C. rodentium and the nleA deletion mutant were mixed, and 100 µl of the mixture was used to infect mice by oral gavage. At the indicated times after infection, mice were sacrificed, and colonic tissues plus faecal pellets were homogenized in PBS using a Polytron tissue homogenizer and serially diluted before being plated on MacConkey agar. After overnight incubation, individual colonies were patched out onto LB plates and, the following day, each clone was genotyped at nleA by PCR using primers flanking the deleted region of the nleA mutant strain. Thirty (C3H/HeJ) or 50 (NIH Swiss) colonies were analysed per mouse, and three mice were analysed per experiment. PCR products were visualized on 1% agarose gels containing ethidium bromide. The competitive index was then determined as the output ratio (mutant:wild type) divided by the input ratio (mutant:wild type).
We would like to thank Richard Pfuetzner for providing purified protein for the generation of antiserum, Carrie Rosenberger for technical advice, Dr Matt Ramer (University of British Columbia) for the use of his microscope for histology, Dr Marilyn Farquhar (UCSD) for providing mannosidase II antiserum, and members of the Finlay laboratory for helpful comments and suggestions on the manuscript. S.G. is supported by a senior research fellowship from the CIHR, N.A.T. by postdoctoral fellowships from NSERC and MSFHR, B.B.F. is the Peter Wall distinguished professor and an HHMI international scholar. Work in the Finlay laboratory is supported by grants from CIHR, HHMI and NSERC.
Note added in proof
While this manuscript was in press, Marches et al.[(2003) Mol Microbiol50: 1553–1567] described another non-LEE encoded type III translocated effector present in some A/E pathogens.