Yersinia YopE is targeted for type III secretion by N-terminal, not mRNA, signals

Authors


Abstract

Pathogenic Yersinia species inject virulence proteins, known as Yops, into the cytosol of eukaryotic cells. The injection of Yops is mediated via a type III secretion system. Previous studies have suggested that YopE is targeted for secretion by two signals. One is mediated by its cognate chaperone YerA, whereas the other consists of either the 5′ end of yopE mRNA or the N-terminus of YopE. In order to characterize the YopE N-terminal/5′ mRNA secretion signal, the first 11 codons of yopE were systematically mutagenized. Frameshift mutations, which completely alter the amino acid sequence of residues 2–11 but leave the mRNA sequence essentially intact, drastically reduce the secretion of YopE in a yerA mutant. In contrast, a mutation that alters the yopE mRNA sequence, while leaving the amino acid sequence of YopE unchanged, does not impair the secretion of YopE. Therefore, the N-terminus of YopE, and not the 5′ end of yopE mRNA, serves as a targeting signal for type III secretion. In addition, the chaperone YerA can target YopE for type III secretion in the absence of a functional N-terminal signal. Mutational analysis of the YopE N-terminus revealed that a synthetic amphipathic sequence of eight residues is sufficient to serve as a targeting signal. YopE is also secreted rapidly upon a shift to secretion-permissive conditions. This ‘rapid secretion’ of YopE does not require de novo protein synthesis and is dependent upon YerA. Furthermore, this burst of YopE secretion can induce a cytotoxic response in infected HeLa cells.

Introduction

The genus Yersinia has three members that are mammalian pathogens. Yersinia pestis is the causative agent of plague, whereas Yersinia pseudotuberculosis and Yersinia enterocolitica are enteric species that cause gastroenteritis (for reviews, see Cornelis and Wolf-Watz, 1997; Cornelis et al., 1998). Although these pathogens cause diseases that differ in nature and severity, they share an ≈ 70 kb plasmid that is required for virulence (Gemski et al., 1980; Zink et al., 1980; Portnoy et al., 1984). The plasmid encodes virulence effector proteins, referred to as Yops (Yersiniaouter proteins), as well as a type III secretion system that mediates the injection of Yops into the cytosol of eukaryotic cells (Cornelis and Wolf-Watz, 1997; Cornelis et al., 1998; Hueck, 1998). The type III secretion system is ‘contact dependent’, in that the secretion of Yops is only triggered when the bacteria come into contact with eukaryotic cells (Rosqvist et al., 1994; Pettersson et al., 1996). Contact can be mimicked in vitro by growing the bacteria in calcium-depleted media, which results in a massive secretion of Yops into the extracellular milieu (Heesemann et al., 1984; Michiels et al., 1990).

The type III secretion system consists of approximately 25 ysc (Yersiniasecretion) genes (Michiels et al., 1991; Plano et al., 1991; Haddix and Straley, 1992; Bergman et al., 1994; Fields et al., 1994; Allaoui et al., 1995; Iriarte and Cornelis, 1999). Ten of these genes (yscCJLNQRSTUV) are conserved in the type III secretion systems of a wide variety of Gram-negative plant and animal pathogens (Hueck, 1998). All these genes, with the exception of yscC, are also conserved in the type III export systems that assemble bacterial flagella (Macnab, 1996). In fact, the Salmonella inv/spa (Kubori et al., 1998) and Shigella mxi/spa (Tamano et al., 2000) type III secretion systems form needle-like structures that resemble a flagellar basal body. A similar structure has yet to be observed in Yersinia, but the LcrV (Pettersson et al., 1999), YopN (Forsberg et al., 1991) and TyeA (Iriarte et al., 1998) proteins have been shown to be surface localized before eukaryotic cell contact. YopN and TyeA presumably serve to cap the type III secretion system, as yopN (Forsberg et al., 1991) and tyeA (Iriarte et al., 1998) mutants constitutively secrete Yops irrespective of eukaryotic cell contact or the presence of calcium in the environment. LcrV forms a pore in the eukaryotic cell membrane (Holmström, 1999) and is required, along with YopB (Sory and Cornelis, 1994; Håkansson et al., 1996; Holmström, 1999) and YopD (Rosqvist et al., 1991; Sory and Cornelis, 1994; Francis and Wolf-Watz, 1998; Holmström, 1999), for the translocation of Yops into eukaryotic cells.

The Yop effectors consist of YopE, YopH, YopJ/YopP, YopM, YopT and YpkA/YopO. YopE has a Rho-GTPase-activating activity (Von Pawel-Rammingen et al., 2000) that, along with YopT (Iriarte and Cornelis, 1998), depolymerizes the actin cytoskeleton of host cells (Rosqvist et al., 1991). It must be noted, however, that YopT is not expressed in the strain used in our study, Y. pseudotuberculosis YPIII(pIB102). YopH is a tyrosine phosphatase (Guan and Dixon, 1990) that disrupts focal adhesions by targeting p130Cas (Black and Bliska, 1997; Persson et al., 1997). Together, the actions of YopE (Rosqvist et al., 1990) and YopH (Rosqvist et al., 1988) prevent Yersinia from being internalized by phagocytes, thereby allowing the bacteria to replicate extracellularly. YopJ induces apoptosis in macrophages (Mills et al., 1997; Monack et al., 1997) and downregulates the NF-κB-mediated inflammatory response (Schesser et al., 1998) by interfering with the MAP kinase signalling pathway (Palmer et al., 1998; 1999; Orth et al., 1999). It has been shown recently that the net result of YopJ activity is the inhibition of phosphorylation of the CREB transcriptional activator (Meijer et al., 2000). YpkA is a serine/threonine kinase (Galyov et al., 1993) that targets the Rho-GTPase signal transduction pathway (Dukuzumuremyi et al., 2000), whereas YopM is a leucine-rich repeat protein that is targeted to the nucleus (Skrzypek et al., 1998), but whose function is unknown.

Signals that target Yops to the type III secretion apparatus have been localized to the N-terminal ends of the Yop proteins. It has been shown that fusing the N-terminal 15 or 17 residues of YopE or YopH, respectively, to the Bordetella pertussis adenlyate cyclase reporter results in the type III-mediated secretion of the hybrid proteins (Sory et al., 1995; Schesser et al., 1996). The N-termini of the Yops do not share a consensus sequence, however, and the manner in which the type III secretion system can recognize such diverse substrates is unclear.

Recently, Schneewind and coworkers, using the neomycin phosphotransferase (Npt) reporter, have accumulated results that suggest that the Yersinia type III secretion signal does not reside in the N-termini of the Yop proteins, but rather in the 5′ end of yop mRNAs. Specifically, they showed that fusing the first 10 or 15 codons of yopE, yopN (Anderson and Schneewind, 1997) or yopK (yopQ in Y. enterocolitica) (Anderson and Schneewind, 1999) to Npt was sufficient to direct the secretion of the reporter. Surprisingly, these secretion signals can, in most cases, be altered by frameshift mutations without affecting the secretion of Npt, which suggests that the amino acid sequences of the Yops are dispensable for targeting. They propose that the 5′ end of yop mRNAs form stem–loop structures that are recognized by the type III secretion system. In their model, interactions of yop mRNA secondary structures with the type III secretion system serve to couple secretion and translation. However, it should be noted that the 5′ mRNA secretion signal is controversial, as a mutational analysis of the YopE secretion signal demonstrated that the first seven amino acids of YopE are important for secretion (Schesser et al., 1996).

In addition, a second secretion signal has been found for YopE (Cheng et al., 1997). This signal consists of amino acids 15–100, which are required for YopE to bind to its cognate chaperone YerA, also known as SycE (Forsberg and Wolf-Watz, 1990; Wattiau and Cornelis, 1993; Schesser et al., 1996; Cheng et al., 1997). Deletion of the yopE mRNA signal did not abolish the secretion of YopE–Npt fusions that contained the YerA binding site (Cheng et al., 1997). In this model, the YerA chaperone is proposed to target YopE to secretion loci by interacting directly with the type III secretion system.

Here, we show that frameshift mutations that alter the first 11 residues of Y. pseudotuberculosis YopE drastically reduce the secretion of YopE in a yerA mutant. In contrast, a mutation that alters the yopE mRNA sequence, without changing the amino acid sequence of YopE, does not impair the secretion of YopE. Furthermore, we demonstrate that a synthetic amphipathic sequence at the YopE N-terminus is sufficient for type III targeting in a yerA mutant. In addition, we show that YopE is rapidly secreted upon a shift to secretion-permissive conditions and that this ‘rapid secretion’ of YopE does not require de novo protein synthesis, but is entirely dependent upon YerA.

Results

Effects of mutations on the secretion of YopE by wild-type and yerA strains

To characterize the Yersinia N-terminal/5′ mRNA type III secretion signal, we focused upon the secretion of the native substrate YopE. YopE was expressed from plasmid pSL155 (see Table 1) such that the untranslated RNA sequences, including the Shine–Dalgarno site, were vector derived. A similar strategy has been used by Schneewind and colleagues in previous analyses of type III secretion signals (Anderson et al., 1999; Anderson and Schneewind, 1999). Expression of YopE was under the tight control of the arabinose-inducible PBAD promoter (Guzman et al., 1995). YopE was expressed in strain YPIII(pIB522), a yopE null strain, at 37°C in a calcium-depleted medium in order to induce Yop secretion. Using this system, YopE was readily secreted into the extracellular milieu (Fig. 1A). As a control, we constructed pSL153, in which yopE codons 2–13 are deleted. Therefore, the N-terminal/5′ mRNA secretion signal is absent. This deletion completely abolished the secretion of YopE (Fig. 1A). Next, we introduced frameshift mutations into the 5′ end of yopE. Frameshift mutations were introduced immediately after the initiation codon, and compensatory mutations were introduced after codon 11 in order to restore the reading frame, thereby allowing the expression of the full-length YopE protein (Fig. 1B). Plasmids pSL161 (−1 frameshift), pSL162 (+1 frameshift), pSL163 (−2 frameshift) and pSL164 (+2 frameshift) were then introduced into strain YPIII(pIB522). All four frameshifted YopE proteins were readily secreted, although in most cases at somewhat reduced levels when compared with wild-type YopE (Fig. 1A). A central tenet, albeit an implicit one, of the 5′ mRNA targeting signal hypothesis is that mutations that alter the sequence, and hence the secondary structure, of the 5′ end of yopE mRNA should result in an uncoupled phenotype in which YopE is expressed, but is unable to be secreted. To test this, we constructed a mutant in which we altered the yopE 5′ mRNA sequence as much as possible, without changing the amino acid sequence of YopE (Fig. 1B and C). To do this, we mutated serine codons 4, 5, 8 and 10 from TCT/A to AGC, which also encodes serine. In addition, the wobble positions of codons 2, 3, 6, 7 and 9 were also mutated. The net effect of these mutations is that 17 out of the 27 nucleotides of codons 2–10 are altered. This plasmid, pSL173, was then introduced into YPIII(pIB522). Secretion of this YopE mutant, denoted mRNAmut, is equivalent to that of wild-type YopE (Fig. 1A, mRNAmut).

Table 1. Strains and plasmids used in this study.
Strain or plasmidRelevant characteristicsReference
E. coli
 DH5αRecipient for cloning experiments Hanahan (1983)
 S17-1λpirRecipient for suicide vector pDM4 Simon et al. (1983)
Yersinia
 YPIII(pIB102)Wild-type, parent strain Bölin et al. (1982)
 YPIII(pIB504) yerA Forsberg and Wolf-Watz (1990)
 YPIII(pIB522) yopE Rosqvist et al. (1988)
 YPIII(pIB525) yerA–yopE This study
Plasmids
 pBAD18Contains PBAD promoter Guzman et al. (1995)
 pET-22bT7 overexpression vectorNovagen
 pDM4Suicide vector Milton et al. (1996)
 pSL151YopE Δ2–13 in pET-22bThis study
 pSL153YopE Δ2–13 in pBAD18This study
 pSL154Wild-type YopE leader in pET-22bThis study
 pSL155Wild-type YopE leader in pBAD18This study
 pSL156YopE −1 frameshifted leader in pET-22bThis study
 pSL157YopE +1 frameshifted leader in pET-22bThis study
 pSL158YopE −2 frameshifted leader in pET-22bThis study
 pSL159YopE +2 frameshifted leader in pET-22bThis study
 pSL161YopE −1 frameshifted leader in pBAD18This study
 pSL162YopE +1 frameshifted leader in pBAD18This study
 pSL163YopE −2 frameshifted leader in pBAD18This study
 pSL164YopE +2 frameshifted leader in pBAD18This study
 pSL165YopE 5′ mRNA mutant in pET-22bThis study
 pSL170YopE serine/isoleucine leader in pET-22bThis study
 pSL173YopE 5′ mRNA mutant in pBAD18This study
 pSL177YopE serine/isoleucine leader in pBAD18This study
 pSL184 ΔyerA–yopE in pDM4This study
 pSL193YopE poly serine leader in pET-22bThis study
 pSL194YopE poly isoleucine leader in pET-22bThis study
 pSL195YopE poly serine leader in pBAD18This study
 pSL196YopE poly isoleucine leader in pBAD18This study
Figure 1.

YopE secretion by wild-type Y. pseudotuberculosis and various yopE mutants.

A. yopE mutants were expressed in trans from an arabinose-inducible promoter in yopE, YPIII(pIB522), or yopE–yerA, YPIII(pIB525), strains. Bacteria were grown in BHI medium lacking calcium at 37°C for 2 h. Cultures were centrifuged in order to separate supernatant (S) and cell pellet (P) fractions. Equal percentages of each fraction were analysed by immunoblotting with an anti-YopE antiserum. WT, wild-type; Δ, deletion of codons 2–13; −1, +1, −2, +2 indicate the reading frame of codons 2–11; mRNAmut indicates a mutant in which the mRNA sequence of codons 2–10 was altered, without changing the amino acid sequence of the protein (see C).

B. Amino acid sequences of the first 13 amino acids of YopE.

C. mRNA sequences of wild-type yopE and the mRNAmut mutant. Serine codons 4, 5, 8 and 10 were mutated from TCT/A to AGC, which also encodes serine. In addition, the wobble positions of codons 2, 3, 6, 7 and 9 were mutated. The net effect of these mutations is that 17 out of the 27 nucleotides of codons 2–10 are altered.

The above results do not allow a definitive conclusion as to the nature of the YopE secretion signal, as all the mutants were secreted at significant levels. The fact that YopE can also be targeted for type III secretion via its chaperone YerA (Cheng et al., 1997) suggested that the observed secretion of the YopE mutant proteins could be YerA mediated. To test this, we expressed all the aforementioned YopE mutants in strain YPIII(pIB525), a yerAyopE double null strain, and measured YopE secretion in liquid cultures. As described previously (Forsberg and Wolf-Watz, 1990; Frithz-Lindsten et al., 1995), wild-type YopE was secreted to a significant degree in this background, although at lower levels than those seen in YPIII(pIB522). In contrast, all the frameshift mutations, which scramble the amino acid sequence of the YopE N-terminus, drastically reduced the secretion of YopE in this strain (Fig. 1A). This suggests that the secretion of the frameshifted YopE mutant proteins seen in YPIII(pIB522) was entirely mediated by YerA, and that the N-terminus of YopE is absolutely required for secretion in the absence of its chaperone. Furthermore, the mRNAmut mutant secreted YopE at wild-type levels in the absence of YerA (Fig. 1A). These results demonstrate conclusively that the N-terminus of YopE serves as a type III targeting signal, whereas yopE mRNA does not.

Cytotoxicity of yopE mutants in wild-type and yerA strains

Next, we determined whether the mutant YopE proteins could be injected into HeLa cells. The injection of YopE into HeLa cells induces a cytotoxic response that results in a rounding up of the HeLa cells, which can be visualized using a phase-contrast microscope (Rosqvist et al., 1991). In our hands, this is the most sensitive method for detecting YopE translocation. The injection of mutant YopE into HeLa cells mirrors what was seen in the liquid culture secretion assay. Namely, when HeLa cells were infected with YPIII(pIB522) strains expressing the different YopE mutants, a cytotoxic response was observed in each instance, with the exception of the mutant in which codons 2–13 were deleted (Fig. 2). In contrast, when HeLa cells were infected with YopE mutant-expressing YPIII(pIB525) yerA strains, a cytotoxic response was only observed for wild-type YopE and the YopE mRNAmut that contains the wild-type N-terminal amino acid sequence. As expected, the yopE mutants that did not secrete in the yerA strain were not cytotoxic. These results confirm that the N-terminus of YopE acts as a signal for type III secretion in vivo.

Figure 2.

HeLa cell cytotoxicity after infection with wild-type and yopE mutants. HeLa cells were infected with yopE, YPIII(pIB522), or yopE–yerA, YPIII(pIB525), strains expressing the various yopE mutants (see Fig. 1), and YopE-induced cytotoxicity was monitored. Photographs were taken at the times indicated.

A synthetic amphipathic sequence is sufficient to target YopE for secretion

As the different Yop proteins do not share a consensus sequence at their N-termini, it has been difficult to understand how the Yersinia secretion machinery could recognize them all. However, a closer inspection of Yop N-terminal sequences revealed a few similarities. Notably, the N-termini of the Yops are amphipathic in nature, which is in contrast to the hydrophobic sequences that target proteins to the general secretory, or sec-dependent, pathway (Pugsley, 1993). There does not seem to be a strong preference for any particular polar amino acid at the Yop N-termini. Among hydrophobic residues, however, there seems to be a preference for isoleucine and, to a lesser extent, leucine. However, the particular N-terminal arrangement of polar and hydrophobic residues varies among the different Yops. Based on this limited amount of data, we replaced codons 2–8 of YopE with heterologous amino acid sequences that differ in their physical properties (Fig. 3). In particular, codons 2–8 were replaced with poly serine (pSL195), poly isoleucine (pSL196) or a synthetic amphipathic sequence consisting of alternating serine and isoleucine residues (pSL177). The different mutants were then expressed in both the YPIII(pIB522) and YPIII(pIB525) strains, and the level of YopE secretion was measured in liquid cultures. Replacement of the YopE N-terminus with either hydrophilic (serine) or hydrophobic (isoleucine) residues drastically reduced the secretion of YopE, in both the presence and the absence of YerA (Fig. 3). Interestingly, the synthetic serine/isoleucine sequence targeted YopE for secretion even in the absence of YerA. These results demonstrate that an amphipathic sequence is capable of serving as a type III targeting signal in Yersinia.

Figure 3.

A synthetic, amphipathic N-terminal sequence is sufficient for YopE secretion.

A. Amino acid residues 2–8 of YopE were replaced with poly serine, poly isoleucine or alternating serine and isoleucine residues (poly S/I).

B. Effects of synthetic leader sequences on YopE secretion. Mutants were expressed in trans from an arabinose-inducible promoter in yopE, YPIII(pIB522), or yopE–yerA, YPIII(pIB525), strains. Bacteria were grown in BHI medium lacking calcium at 37°C for 2 h. Cultures were centrifuged in order to separate supernatant (S) and cell pellet (P) fractions. Immunoblots were performed on equal percentages of each fraction using an anti-YopE antiserum.

Secretion kinetics of YopE

The fact that there are two ways in which YopE can be targeted for secretion is interesting and suggests that each targeting signal may have a distinct role during an infection. In order to address this question, we examined the kinetics of YopE secretion in liquid cultures after shifting the bacteria from secretion-inhibiting (+Ca2+) to secretion-permissive (–Ca2+) conditions (Michiels et al., 1990). Specifically, both wild-type, YPIII(pIB102), and yerA null, YPIII(pIB504), strains were grown at 37°C for 1 h in a calcium-containing medium. A sample of each culture was removed (0 min time point); the calcium chelator EGTA was then added, and the amount of YopE secreted into the extracellular medium was measured at various time points. YopE was secreted immediately by the wild-type strain, whereas secretion was not observed until 30 min after induction for the yerA strain (Fig. 4).

Figure 4.

Kinetics of YopE secretion after calcium depletion. Wild-type, YPIII(pIB102), and yerA, YPIII(pIB504), strains were grown at 37°C for 1 h in BHI medium containing 2.5 mM calcium. At time 0, a 1 ml sample was removed, followed by the addition of EGTA to a final concentration of 10 mM. At the times indicated, 1 ml samples were centrifuged in order to separate supernatant (S) and cell pellet (P) fractions. Immunoblots were performed on equal percentages of each fraction using an anti-YopE antiserum.

The fact that YopE was detected in both wild-type and yerA strains before the addition of EGTA demonstrates that the translation of YopE is not coupled to secretion. To confirm this, we examined whether the rapid YopE secretion seen in the wild-type strain was dependent upon de novo protein synthesis. Rapid secretion of YopE by the wild-type strain occurred even in the presence of chloramphenicol. In contrast, no YopE secretion was observed under the same conditions in the yerA mutant (Fig. 5). This suggests that the rapid burst of YopE secretion seen with the wild-type strain does not require de novo protein synthesis, but is absolutely dependent upon YerA. Finally, we tested whether or not this protein synthesis-independent burst of YopE secretion was sufficient to induce a cytotoxic response in HeLa cells. Specifically, HeLa cells were infected with either the wild-type strain, YPIII(pIB102), or a yerA mutant strain, YPIII(pIB504), in the presence of 100 µg ml−1 chloramphenicol. The short burst of YerA-mediated YopE secretion that occurs in the absence of protein synthesis is sufficient to induce a cytotoxic response in infected HeLa cells (Fig. 6), indicating that these early events in host–microbe interactions are likely to be very important during an infection.

Figure 5.

Protein synthesis-independent YopE secretion. Wild-type, YPIII(pIB102), and yerA, YPIII(pIB504), strains were grown for 1 h at 37°C in BHI medium containing 2.5 mM calcium. Chloramphenicol was added to 100 µg ml−1. Fifteen minutes later, the bacteria were pelleted and resuspended in BHI medium containing EGTA. YopE levels in the supernatant (S) and cell pellet (P) fractions were measured using an anti-YopE antiserum.

Figure 6.

YopE cytotoxicity in the presence of chloramphenicol is YerA dependent. HeLa cells were infected with wild-type, YPIII(pIB102), or yerA, YPIII(pIB504), strains in the presence of chloramphenicol. Chloramphenicol (100 µg ml−1) was added to both the cell culture medium and the bacteria 15 min before infection. Photographs were taken 1 h after infection.

Discussion

YopE secretion signals were first identified through the use of the Bordetella pertussis adenylate cyclase (Cya) reporter (Sory et al., 1995). It was shown that fusing as few as the N-terminal 11 amino acids of YopE to Cya was sufficient for the type III-dependent export of the hybrid protein. Similar results were also obtained using the N-terminus of YopH (Sory et al., 1995; Schesser et al., 1996). These results led to the view that the N-termini of the Yops serve as targeting signals for type III secretion. The N-termini of the Yops do not, however, share a consensus sequence, and the manner in which such diverse sequences could be recognized by the secretion apparatus has remained unclear.

More recently, Schneewind and coworkers have proposed that type III targeting signals do not consist of the N-termini of the Yops, but rather the 5′ end of yop mRNAs. Specifically, they have demonstrated that fusing the N-terminal 10 or 15 residues of YopE, YopN (Anderson and Schneewind, 1997) or YopK (YopQ in Y. enterocolitica) (Anderson and Schneewind, 1999) to the neomycin phosphotransferase (Npt) reporter is sufficient to direct the type III-dependent export of the hybrid proteins. In addition, they showed that the introduction of frameshift mutations into the Yop targeting sequences did not interfere with secretion. This led them to propose that the amino acid sequences of the Yops are dispensable for secretion and that it is the yop mRNAs that serve as targeting signals. In their model, yop mRNAs are predicted to fold into stem–loop structures in which the Shine–Dalgarno and AUG initiation sequences are sequestered into a stem, thereby inhibiting translation. Upon interaction of these stem–loop structures with a component of the secretion apparatus, the translational inhibition is relieved, thereby coupling Yop translation and secretion. Furthermore, mRNA-mediated type III targeting is thought to be universal, as 5′ mRNA sequences are sufficient to target the AvrB and AvrPto virulence proteins of the plant pathogen Pseudomonas syringae for type III secretion (Anderson et al., 1999).

It must be noted, however, that the mRNA targeting model has several shortcomings. First, there is no evidence that yop mRNAs form stem–loop structures, either in vivo or in vitro. Secondly, there is no biochemical evidence that suggests that yop mRNAs interact with any component of the secretion apparatus. Thirdly, an mRNA targeting signal has never been demonstrated for a native substrate of any type III secretion system. Fourthly, the Npt reporter used by Schneewind and colleagues is exported in a targeting signal-independent manner by the flagellar type III secretion system, and a mutational analysis of the secreted flagellar protein FlgM failed to provide evidence in support of an mRNA targeting signal (Chilcott and Hughes, 1998). In particular, fusing random, vector-derived sequences to the N-terminus of NPT was sufficient to mediate the type III-dependent export of this reporter by the flagellar export apparatus. Also, all randomly isolated, secretion-impaired FlgM mutants have amino acid substitutions at their N-termini. No mutations were found in the wobble positions at the 5′ end of the flgM coding sequence, as would be predicted if flgM mRNA serves as a targeting signal. Fifthly, direct interactions between exported substrates and components of the secretion apparatus have been demonstrated in the flagellar system, which suggests that protein–protein interactions may be sufficient to mediate type III targeting (Silva-Herzog and Dreyfus, 1999; Minamino and MacNab, 2000). Sixthly, a site-directed mutagenesis study suggested that the first seven amino acids of YopE are important for type III secretion (Schesser et al., 1996). Finally, the cellular levels of YopH in a sycH mutant, in which the cognate YopH chaperone is not expressed, is equivalent to that of the wild-type strain (Persson et al., 1995). This is despite the fact that the secretion of YopH is almost completely abolished in the sycH mutant. It is difficult to reconcile this fact with the mRNA targeting signal hypothesis, which states that translation and secretion are coupled processes.

In order to determine whether the N-terminus of YopE or the 5′ end of yopE mRNA serves as a type III secretion signal, we systematically mutagenized the first 11 codons of YopE. In our system, the native type III secretion substrate YopE was expressed from a plasmid such that the untranslated RNA sequences, including the Shine–Dalgarno site, were vector derived. Wild-type yopE untranslated RNA sequences are predicted to be required in order for yopE mRNA to form the stem–loop structure that serves as a type III targeting signal. Nevertheless, we easily detected YopE secretion despite the heterologous nature of the untranslated RNA sequences in our system. Schneewind and coworkers have also shown that untranslated RNA sequences are dispensable for type III secretion (Anderson et al., 1999; Anderson and Schneewind, 1999) As a control, we deleted codons 2–13 of yopE. This deletion completely abolished the secretion of YopE. This is in contrast to results obtained using the Npt reporter. Specifically, Schneewind and coworkers showed that deleting the first 15 codons of yopE still allowed the secretion of a YopE–Npt fusion protein, at roughly 15% of the levels seen for wild-type YopE (Cheng et al., 1997).

Next, we introduced frameshift mutations into the N-terminus of YopE. As seen previously, the frameshifted YopE leader sequences sufficed to target YopE for secretion. However, we also showed that YopE was secreted at wild-type levels by a mutant possessing an mRNA sequence that contained alterations at 17 out of the 27 nucleotides of codons 2–10, while still encoding the native amino acid sequence. It has been shown previously that the chaperone YerA can also target YopE for secretion (Cheng et al., 1997). Therefore, we tested whether the aforementioned mutant proteins could be secreted in a strain lacking YerA. Our results show definitively that the secretion of the frameshifted YopE mutant proteins is drastically reduced in the absence of the YerA chaperone. In striking contrast, YopE expressed by the mutant containing the altered mRNA sequence was secreted at levels comparable with that of wild-type YopE. Taken together, these results show conclusively that YopE is targeted for type III secretion by both its N-terminus and the chaperone YerA. We note that the N-termini of exported substrates also serve as targeting signals in the flagellar type III secretion system (Chilcott and Hughes, 1998). As the type III secretion systems of Gram-negative pathogens presumably evolved from the flagellar export apparatus, these results suggest that all type III secretion systems recognize the N-termini of secreted substrates.

Although the N-termini of the Yops do not share a consensus sequence, we note that all Yop N-termini are amphipathic in nature. To test whether such an amphipathic sequence is required for secretion, we replaced the YopE N-terminus with artificial polar, hydrophobic or amphipathic sequences. Only an N-terminal synthetic amphipathic sequence consisting of alternating serine and isoleucine residues was secreted, in both the presence and the absence of YerA. Neither poly serine nor poly isoleucine sequences at the N-terminus of YopE permitted secretion, despite the presence of YerA. These results suggest that the sequence of the YopE N-terminus is important even for the YerA-mediated secretion pathway. Also note that deleting amino acids 2–13 of YopE abolishes YerA-mediated secretion. Although it is known that the N-terminal 15 residues of YopE are dispensable for YerA binding (Woestyn et al., 1996; Cheng et al., 1997), it is possible that the poly serine and poly isoleucine leader sequences inhibit the YopE–YerA interaction. In any event, the failure of the poly serine and poly isoleucine leader sequences to support YopE secretion suggests that an amphipathic sequence at the N-terminus of YopE is a prerequisite for secretion. Further studies will be needed to determine whether other synthetic amphipathic sequences are also capable of targeting Yops for secretion.

The reasons why the N-terminus of YopE is required for YerA-mediated secretion are currently unknown. We speculate that once YerA targets YopE to the secretion apparatus, the YopE–YerA interaction must be abrogated in order for YopE to be secreted. It is possible that the N-terminus of YopE serves as a handle by which a component of the secretion apparatus, possibly the putative ATPase YscN (Woestyn et al., 1994), releases YopE from YerA.

The manner in which the Yersinia secretion machinery recognizes the diverse N-termini of exported substrates is unknown, although there is precedence for such recognition events. The chaperone GroEL, for example, binds approximately 300 substrates in vivo (Houry et al., 1999). These substrates share common secondary structural motifs, but lack significant sequence homology. The crystal structure of GroEL bound to a peptide substrate suggests that unstructured peptides mould themselves into an amphipathic cleft in the GroEL protein (Chen and Sigler, 1999). We have shown that an 11-amino-acid peptide corresponding to the N-terminus of YopE is unstructured in solution (data not shown). This is also true for the N-termini of the secreted flagellar proteins FlgM (Daughdrill et al., 1997), the hook protein FlgE and flagellin (Vonderviszt et al., 1989; 1992). Therefore, as first proposed by Hughes and coworkers, it is possible that unstructured, amphipathic sequences are bound by one or more components of the secretion apparatus (Chilcott and Hughes, 1998).

The fact that YopE can also be targeted for type III secretion by its chaperone YerA is interesting and suggests that this signal may play a unique role during host–microbe interactions. Our investigation of the kinetics of YopE secretion supports such a suggestion. In a wild-type strain, YopE is secreted immediately after shifting the bacteria from secretion-inhibiting to secretion-permissive conditions. In a yerA mutant, secretion of YopE was not observed for 30 min, despite an abundance of YopE in the cell. The fact that there is an accumulation of cytoplasmic YopE in the yerA mutant suggests that the translation of YopE is not coupled to secretion, as is predicted by the mRNA targeting signal hypothesis. In further support of this view, YopE was detected in both wild-type and yerA strains before the induction of secretion.

In addition, we showed that the rapid secretion of YopE by the wild-type strain is not dependent upon de novo protein synthesis, and that this burst of YopE secretion is sufficient to induce a cytotoxic response in infected HeLa cells. Similar results have been obtained previously by several groups (Goguen et al., 1986; Fällman et al., 1995; Black and Bliska, 1997). Here, we demonstrate that this protein synthesis-independent YopE secretion is entirely dependent upon the chaperone YerA. This suggests that there is a premade pool of YopE that is targeted to the secretion apparatus before contact with a eukaryotic cell. Again, this contradicts the mRNA targeting signal hypothesis, which states that translation and secretion are coupled processes.

The ability of Yersinia to secrete YopE rapidly may play an important role during the course of an infection, as a yerA mutant is avirulent in a mouse model when administered by intraperitoneal injection (Rosqvist et al., 1990). Many other Yops also have cognate chaperones (Wattiau et al., 1994; Day and Plano, 1998; Jackson et al., 1998; Iriarte and Cornelis, 1999) that may serve to ensure that Yop secretion occurs rapidly upon eukaryotic cell contact. However, several Yops, including YopJ, YopM and YpkA/YopO, may not have chaperones. It may be that Yops with chaperones are secreted immediately upon eukaryotic cell contact, whereas the secretion of Yops without chaperones is delayed. Further studies will be needed to determine whether Yop secretion is temporally regulated in this manner.

Experimental procedures

Media and growth conditions

The liquid growth medium for Yersinia strains consisted of brain–heart infusion (BHI) broth (Difco) supplemented with either 5 mM EGTA and 20 mM MgCl2 (BHI minus Ca2+) or 2.5 mM CaCl2 (BHI plus Ca2+). Escherichia coli strains were grown in Luria–Bertani broth or on Luria–Bertani agar (Davis et al., 1980). Bacteria with antibiotic resistance were grown in the presence of the appropriate antibiotic(s) at a final concentration of 100 µg ml−1 chloramphenicol, 50 µg ml−1 (kanamycin) or 100 µg ml−1 (carbenicillin).

DNA methods

Preparation of plasmid DNA, restriction enzyme digests, ligations and transformations of E. coli were performed as described by Sambrook et al. (1989). DNA sequencing reactions were performed using the ThermoSequenase dye terminator cycle sequencing kit and analysed with a SEQ4 × 4 sequencer (Amersham Pharmacia Biotech).

Construction of yopE mutants

A yopE fragment consisting of codons 14–220 was made by amplifying, via polymerase chain reaction (PCR), YPIII(pIB102) genomic DNA with Pfx DNA polymerase (Life Technologies). An NdeI site replaced codons 12 and 13, and a HindIII site was introduced after the yopE stop codon. This fragment was digested with NdeI and HindIII and cloned into the same sites of the pET-22b vector to yield pSL151. To construct the various yopE mutants, pSL151 was first digested with NdeI. Oligonucleotide cassettes consisting of codons 2–12 of yopE and containing the mutations of interest were then ligated into the NdeI site. This changes codon 12 from CCC to CCT (still encodes proline) and codon 13 from CTG to ATG (methionine replaces leucine). The yopE constructs were then moved from the pET-22b vector into the pBAD18 vector using the restriction enzymes HindIII and XbaI. This places the YopE mutants under the control of an arabinose-inducible promoter in which the Shine–Dalgarno sequence is from the pET-22b vector. All mutations were confirmed by DNA sequencing. A list of the different oligonucleotides used in this study is available from the authors upon request.

Construction of the yerA–yopE mutant

A yerA–yopE in frame deletion was made by PCR amplifying YPIII(pIB102) genomic DNA with Pfx DNA polymerase and the primer pairs MN36/MN40 (which results in a fragment complementary to the 3′ end of the yerA gene) and MN41/MN42 (which results in a fragment complementary to the 3′ end of the divergently transcribed yopE gene). The two fragments were ligated by PCR with the primer pair MN36/MN42. The resulting PCR product was digested with SphI and XbaI and cloned into the same sites in the suicide vector pDM4. The resulting construct, pSL184, was then transformed into the E. coli strain S17-1λpir (Simon et al., 1983) and conjugated into the wild-type Yersinia strain YPIII(pIB102) by plating on Yersinia agar (Difco) plates containing chloramphenicol. Exconjugates were restreaked on LB plates containing sucrose in order to counterselect against bacteria still containing the pSL184 plasmid. Sucrose-resistant colonies were PCR amplified with the primer pair MN36/MN42 to confirm the presence of the deletion. The resulting strain, pIB525, lacks codons 1–77 of yerA, the intergenic divergent promoter between yerA and yopE and codons 1–168 of yopE.

YopE secretion assay

Overnight cultures of Y. pseudotuberculosis strains YPIII (pIB522) and YPIII(pIB525) expressing the different YopE constructs were grown in BHI minus Ca2+ medium supplemented with carbenicillin (to select for the YopE-expressing constructs) and kanamycin (to select for the virulence plasmid) at 26°C. The cultures were diluted to OD600 of 0.2 into 10 ml of fresh medium and grown at 26°C for 1 h. Arabinose was added to a final concentration of 0.2%, and the cultures were grown for an additional 2 h at 37°C to induce secretion. The cultures (9 ml) were then centrifuged at 3000 g for 15 min. The supernatant containing the secreted Yops was passed through a 0.45 µM filter and precipitated with 10% trichloroacetic acid (TCA). The TCA precipitates were centrifuged at 3000 g for 20 min, the supernatants were discarded, and the remaining pellets were dried at room temperature. The pellets were resuspended in 250 µl of 2% SDS and precipitated with acetone at −20°C for 30 min. The samples were centrifuged at 20 800 g for 10 min, the supernatants were discarded, and the pellets were air dried. The pellets were then resuspended in 100 µl of 8 M urea, and an equal amount of 2× sample buffer was added. The bacterial cell pellets were resuspended in 100 µl of H2O and an equal amount of 2× sample buffer. Equal amounts of culture supernatant and cell pellet fractions were separated by SDS–PAGE and transferred to a nitrocellulose membrane. YopE was detected with a polyclonal anti-YopE antibody.

Cytotoxicity assay

Overnight cultures of YPIII(pIB522) and YPIII(pIB525) expressing the different YopE constructs were grown in LB medium containing carbenicillin and kanamycin. Cytotoxicity was assayed as described previously (Rosqvist et al., 1990). Arabinose (0.2%) was added to the cell culture medium at day 2 in order to induce YopE expression. Pictures were taken with a phase-contrast microscope (Zeiss).

Kinetics of YopE secretion

Overnight cultures of Y. pseudotuberculosis strains YPIII (pIB102) and YPIII(pIB504) were grown in BHI plus Ca2+ medium supplemented with kanamycin (Km) at 26°C. The bacteria were diluted to an OD of 0.2 into fresh medium and grown at 26°C for 1 h. Cultures were then shifted to 37°C and grown for an additional hour. EGTA was added to a final concentration of 10 mM, and MgCl2 was added to a 20 mM final concentration. Samples (1 ml) were withdrawn at various time points. The samples were immediately centrifuged at 20 800 g for 20 s, and the supernatants were passed through a 0.45 µM filter. The detection of secreted YopE was performed as described previously for the secretion assay.

Protein synthesis-independent YopE secretion

Overnight cultures of Y. pseudotuberculosis strains YPIII (pIB102) and YPIII(pIB504) were grown in BHI plus Ca2+ supplemented with kanamycin (Km) at 26°C. The bacteria were diluted to an OD of 0.2 into fresh medium and grown at 26°C for 1 h. Cultures were then shifted to 37°C and grown for an additional 3 h. Chloramphenicol was added to 100 µg ml−1 either 15 or 30 min before harvesting the cultures. The cultures (10 ml) were then centrifuged at 3000 g for 15 min. The bacteria were resuspended in 1 ml of BHI minus Ca2+ medium. The samples were immediately centrifuged at 20 800 g for 20 s, and the supernatants were passed through a 0.45 µM filter. The detection of secreted YopE was performed as described previously for the secretion assay. Cytotoxicity experiments were performed as described above except that 100 µg ml−1 chloramphenicol was added both to the cell culture medium and to the bacteria 15 min before infection.

Immunoblotting

Samples were separated by 12% SDS–PAGE and electroblotted (Trans Blot SD; Bio-Rad) onto a nitrocellulose transfer membrane (Protran; Schleicher and Schuell) using a transfer buffer containing 25 mM Tris, 192 mM glycine and 20% methanol. The membrane was blocked for 1 h with Tris-buffered saline plus 0.1% Tween (TBS-T) and 5% non-fat dry milk. The membrane was probed for 1 h with a polyclonal anti-YopE antiserum in 10 ml of the blocking buffer and was then washed for 3 × 5 min with TBS-T. The membrane was incubated for 1 h with an anti-rabbit antibody (Amersham Pharmacia Biotech) in 10 ml of blocking buffer, followed by washing with TBS-T. The proteins were detected using the ECL detection kit (Amersham Pharmacia Biotech) according to the manufacturer's instructions.

Acknowledgements

This work was supported by the Swedish Medical Research Council and the Swedish Foundation of Strategic Research.

Ancillary