Molecular and functional analysis of the type III secretion signal of the Salmonella enterica InvJ protein

Authors


Summary

Central to the pathogenicity of Salmonella enterica is the function of a type III secretion system (TTSS) encoded within a pathogenicity island at centisome 63 (SPI-1). An essential component of this system is a supramolecular structure termed the needle complex. Proteins to be delivered into host cells possess specific signals that route them to the type III secretion pathway. In addition, some bacterial proteins have signals that deliver them to the secretion complex to either become their structural components or exert their function at that location. One of these proteins is InvJ, which controls the length of the needle substructure of the needle complex. In this study, we have analysed the signal that targets InvJ to the TTSS. We found that amino acid residues 4 to 7 of InvJ are necessary and sufficient to mediate secretion of InvJ or a reporter protein in a TTSS-dependent manner. InvJ secretion was found to be essential for its function in needle length determination, effector protein secretion and bacterial invasion of epithelial cells. Frameshift mutagenesis analysis indicated that the InvJ type III secretion signal sequence tolerates significant alterations in its amino acid sequence without affecting InvJ secretion. Introduction of silent mutations in the secretion signal coding sequence that result in drastically different predicted mRNA folds had no effect on InvJ secretion or expression.

Introduction

Several important Gram-negative bacteria that are pathogenic for animals and plants have evolved a specialized protein secretion system termed type III (TTSS) (Hueck, 1998; Galán and Collmer, 1999; Cornelis and Van Gijsegem, 2000). This system has evolved to mediate the transfer of proteins from the bacterial cytoplasm into eukaryotic cells, where they modulate cellular functions. Made up of more than 20 proteins and evolutionary related to the flagellar apparatus, TTSSs are among the most complex protein secretion systems known. The enteric pathogenic bacterium Salmonella enterica encodes two TTSSs, which are required for different stages of the pathogenic cycle (Galán, 2001). One of these systems, which is encoded within a pathogenicity island (SPI-1) at centisome 63, mediates the initial interactions of Salmonella with intestinal epithelial cells that lead to bacterial internalization and the production of pro-inflammatory cytokines (Galán, 2001).

A central component of TTSSs is a supramolecular structure termed the needle complex (Kubori et al., 1998). Embedded within the bacterial envelope, this structure is composed of a multiring base that is anchored to the inner and outer membranes and a needle-like projection that spans ≈ 45 nm from the bacterial surface. In addition to the needle complex, there are several highly conserved inner membrane proteins and an ATPase, all of which are essential for secretion. Assembly of the needle complex occurs in a series of discrete steps that can be dissected genetically (Kubori et al., 2000; Sukhan et al., 2001). Assembly of the base requires the sec pathway and is followed by the assembly of the needle portion, a process that is sec independent and requires a functional TTSS. The actual length of the needle substructure is determined by a mechanism that involves the TTSS-associated protein InvJ (Kubori et al., 2000). A loss-of-function mutation in invJ results in a needle complex that possesses an abnormally long needle substructure and a TTSS that is unable to secrete any protein other than those associated with the needle component (Kubori et al., 2000). Therefore, InvJ is required for the TTSS translocon to switch specificity of substrates from those associated with the assembly of the needle complex to those destined to be translocated into host cells. As a consequence, invJ mutants are unable to stimulate cellular responses and are therefore unable to enter host cells. The mechanisms by which InvJ carries out its needle length control function are not understood. InvJ itself can be engaged and secreted via the type III secretion machinery (Collazo et al., 1995), but it is not known how this process relates to its function.

Proteins destined to travel through the type III secretion pathway possess discrete signals that target them to the secretion apparatus (Hueck, 1998; Galán and Collmer, 1999; Cornelis and Van Gijsegem, 2000; Aldridge and Hughes, 2001; Lloyd et al., 2001a). Earlier work carried out by Cornelis and collaborators indicated that the secretion signals of at least some type III secreted proteins were located at the amino-terminus of the secreted polypeptide (Michiels and Cornelis, 1991; Sory and Cornelis, 1994; Sory et al., 1995). This notion was later challenged by the hypothesis that the targeting signal was encoded by the messenger RNA of the secreted protein rather than by its amino acid sequence (Anderson and Schneewind, 1997; 1999; Anderson et al., 1999). This hypothesis was based on the finding that frameshift mutations in the initial codons of the messenger RNA of the secreted protein did not affect the secretion of a fused reporter protein. However, more recent studies have questioned this hypothesis. Wolf-Watz and collaborators found that frameshifting the initial codons of the mRNA of the Yersinia TTSS secreted protein YopE abolished its ability to mediate secretion when these experiments were carried out in the context of the native YopE protein and in the absence of its cognate chaperone (Lloyd et al., 2001b). These studies led to the conclusion that the secretion signal is encoded by the polypeptide sequence but not by the mRNA, and suggested that amphipathic sequences at the amino-terminus might favour the secretion process. The reasons for the contradictory results are unclear, although they may result from differences in the experimental conditions.

In addition to the amino-terminal signal sequence, another very important component of the secretion system is a family of customized chaperones that bind specifically to cognate effector proteins controlling their secretion and, in some cases, their stability and expression (Wattiau et al., 1996; Cornelis and Van Gijsegem, 2000; Lloyd et al., 2001a). Although quite diverse at the primary amino acid sequence level, recent crystallographic studies have shown that these chaperones share a rather strikingly similar tertiary structure (Birtalan and Ghosh, 2001; Luo et al., 2001; Stebbins and Galán, 2001; Evdokimov et al., 2002). In the virulence-associated TTSSs, these chaperones bind a discrete domain of the cognate secreted proteins that is usually located at the amino-terminus. In contrast, chaperones associated with the flagellar TTSS bind a domain located at the carboxyl-terminus of the target protein (Bennett et al., 2001). Most chaperones are specific for a single protein, although some chaperones facilitate the secretion of more than one protein. The recent determination of the crystal structure of one of these chaperone–effector protein complexes (Stebbins and Galán, 2001) established that the function of TTSS-associated chaperones may be to maintain the target proteins in a secretion-competent state. This state would be one of an unfolded polypeptide with secondary structural features that may be essential for their recognition by the secretion machinery.

Although most proteins travelling the type III secretion pathway are destined to be delivered to eukaryotic cells, a subset of these proteins is not injected into cells. Some secreted proteins of S. enterica are structural components of the needle complex (e.g. PrgI) or required for secretion (e.g. InvJ) and, therefore, are considered to be components of the secretion system (Collazo et al., 1995; Kubori et al., 2000). It is not known whether this different class of type III-secreted proteins is targeted to the secretion machinery by different mechanisms. No chaperones that assist the secretion of these proteins have been yet identified, and it is possible that the secretion of these proteins is indeed chaperone independent. To gain insight into the mechanisms of secretion of these types of proteins, we have carried out a detailed analysis of the secretion signal of the TTSS-associated protein InvJ. Our studies indicate that the InvJ secretion signal is contained within its first seven amino acids. Although our studies demonstrate a remarkable tolerance for changes in the primary amino acid sequence of the secretion signal, we could not im-plicate the mRNA encoding the signal sequence in the secretion process. We propose a model to account for the apparent conundrum of high specificity of secretion coupled with a high degree of flexibility in the primary amino acid sequence of the signal sequence.

Results

Identification of the InvJ secretion signal

A number of studies have established that the secretion signals that target proteins destined to be delivered into host cells by the TTSS are localized at their amino-terminus (Galán and Collmer, 1999; Cornelis and Van Gijsegem, 2000; Lloyd et al., 2001a). However, no studies have been carried out to define the secretion signal of proteins that, although secreted via the TTSS, are components of the secretion system. We therefore carried out a deletion analysis to define the minimum domain of InvJ that is capable of mediating type III secretion of a reporter protein. We constructed plasmids encoding chimeric proteins composed of either 179 or 52 residues of the amino-terminus of InvJ fused to the Escherichia coli alkaline phosphatase (PhoA) lacking its native secretion signal sequence. The different plasmids were introduced into wild-type S. enterica serovar Typhimurium (S. typhimurium) and an isogenic strain defective in type III secre-tion resulting from a non-polar mutation in invA, which encodes an essential component of this system (Galán et al., 1992). Whole-cell lysates and culture supernatants of the resulting strains were examined by Western blot for the expression and secretion of the hybrid proteins. Both the first 179 and 52 amino-terminal residues of InvJ were able to mediate the secretion of the reporter protein PhoA efficiently (Fig. 1A). These results indicated that the InvJ secretion signal is present within a discrete domain of its amino-terminus, in close agreement with what has been observed in type III secreted proteins that are destined to be translocated into host cells (effector proteins).

Figure 1.

A. The amino-terminus of InvJ directs the secretion of alkaline phosphatase. The first 179 (pSB485) or 52 (pSB778) amino acids of InvJ were fused to alkaline phosphatase (PhoA) devoid of its secretion signal and expressed in either wild-type (SB300) or invA (SB136) S. typhimurium. Whole-cell lysates and culture supernatants of the resulting strains were analysed for the presence of alkaline phosphatase by western immunoblot as described in Experimental procedures.

B. The first seven amino acids of InvJ are sufficient to direct the secretion of alkaline phosphatase. The first 25, 15, 11, 8, 7, 6 or 5 amino acids of InvJ were fused to alkaline phosphatase (PhoA) devoid of its secretion signal and expressed in wild-type S. typhimurium. Whole-cell lysates and culture supernatants of the resulting strains were analysed for the presence of alkaline phosphatase by Western immunoblot as described in Experimental procedures.

To delineate the InvJ secretion signal further, we constructed a series of chimeric proteins composed of increasingly smaller numbers of residues from the amino-terminus of InvJ fused to PhoA. Chimeric proteins containing the first 25, 15, 11, eight and seven amino-terminal residues of InvJ fused to PhoA were secreted efficiently in a type III secretion-dependent manner by S. typhimurium (Fig. 1B). In contrast, the first five or six amino acids of InvJ were unable to mediate the secretion of PhoA (Fig. 1B). The levels of all the chimeric proteins in whole-cell lysates were equivalent, indicating that the absence of InvJ1–5-PhoA and InvJ1–6-PhoA from culture supernatants resulted from lack of secretion and not lack of expression of these chimeric proteins (Fig. 1B). These results indicate that the secretion signal of InvJ is encoded within its first seven amino acids (or codons) and that this signal is sufficient to mediate the secretion of a reporter protein.

To delineate the amino-terminal boundary of the secretion signal of InvJ, we progressively introduced single amino acid deletions at the amino-terminus. Deletion of the second (InvJΔ2–PhoA) as well as the second and third (InvJΔ2−3–PhoA) amino acids from the 7-amino-acid secretion signal did not affect the secretion of the PhoA reporter protein (Fig. 2). In contrast, deletion of amino acids 2–4 (InvJΔ2−4–PhoA) or 2–5 (InvJΔ2−5–PhoA) completely abolished the secretion of the InvJ–PhoA chimeric proteins (Fig. 2). Expression of the InvJΔ2−5–PhoA chimeric protein was reduced in comparison with the other constructs, although such a reduction in expression cannot account for its complete absence from the culture supernatant. Taken together, these results established the importance of the InvJ residues (or codons) 4–7 in mediating the secretion of the PhoA reporter protein.

Figure 2.

Delineation of the amino-terminal boundary of the InvJ secretion signal. Chimeric proteins consisting of different deletions of the InvJ secretion signal and alkaline phosphatase (PhoA) were expressed in wild-type S. typhimurium. Whole-cell lysates and culture supernatants of the resulting strains were analysed for the presence of alkaline phosphatase by Western immunoblot as described in Experimental procedures.

We investigated whether the same InvJ residues (or codons) shown to be essential for the type III secretion of the reporter PhoA protein were also essential for InvJ secretion in the context of the full-length protein. For this purpose, we constructed a series of amino-terminal InvJ deletions and examined their secretion in an invJ S. typhimurium mutant strain. Consistent with the results obtained with the PhoA reporter protein, removal of amino acids 2 and 3 of InvJ did not affect its secretion and/or expression (Fig. 3). Removal of any more amino acids from the N-terminus of InvJ effectively blocked its secretion (Fig. 3). Expression of the InvJ deletion constructs InvJΔ2−4, InvJΔ2−5 and InvJΔ2−6 was reduced in comparison with the wild-type protein. However, the reduction in expression of these constructs cannot account for their complete absence from culture supernatants as their expression was equivalent to that of InvJΔ2 and InvJΔ2−3, which were efficiently secreted and readily detected in culture supernatants (Fig. 3). Furthermore, expression of InvJΔ2−7 was indistinguishable from that of the wild-type protein, although this construct was not detected in culture supernatants. These results indicate that the fourth, fifth, sixth and seventh codons of InvJ encode information essential for its secretion.

Figure 3.

Delineation of the InvJ secretion signal in the context of the full-length protein. InvJ mutants carrying various deletions at the amino-terminus as indicated were expressed in an isogenic S. typhimurium invJ mutant strain. Whole-cell lysates and culture supernatants of the resulting strains were analysed for the presence of InvJ by Western immunoblot as described in Experimental procedures.

The InvJ secretion signal tolerates significant changes in the coding sequence

Analysis of the secretion signal of translocated effector proteins of Yersinia spp. has led to seemingly contradictory conclusions regarding the nature of the type III secretion signal. Although experiments in some laboratories have favoured the hypothesis that, at least in some cases, the type III secretion signal resides within the mRNA of the coding sequence (Anderson and Schneewind, 1997; 1999), other studies have shown evidence that the secretion signal resides within the amino acid sequence (Michiels and Cornelis, 1991; Sory et al., 1995; Lloyd et al., 2001b; 2002). We therefore carried out experiments to investigate whether the secretion signal of InvJ, a type III secreted protein that is not translocated into host cells, was determined by its amino acid or mRNA sequence. We first introduced +1 or −1 reading frameshift mutations by either introducing or deleting a guanine nucleotide immediately downstream of the ATG initiation codon of the efficiently secreted InvJ1−15–PhoA fusion protein. Although both frameshift mutations resulted in completely different predicted amino acid sequence, these changes did not alter either the secretion or the expression of the InvJ–PhoA chimeric proteins in the different strains (Fig. 4). Similar results were obtained when −1 or +1 reading frameshifts were introduced immediately downstream of the ATG initiation codon of the InvJ1−7–PhoA chimeric protein, which is efficiently secreted via the TTSS (Fig. 4; data not shown). Frameshifting the non-secreted InvJ1−6–PhoA chimeric protein did not restore its secretion (Fig. 4).

Figure 4.

Introduction of frameshift mutations in the secretion signal of InvJ does not alter its ability to direct the secretion of alkaline phosphatase. Frameshift mutations (+1 or −1) were introduced in the first seven or 15 amino acids of InvJ fused to alkaline phosphatase and expressed in wild-type S. typhimurium. Whole-cell lysates and culture supernatants of the resulting strains were analysed for the presence of alkaline phosphatase by Western immunoblot as described in Experimental procedures.

It has been argued that some of the discrepancy in the results obtained by different laboratories may have resulted from the use of reporter proteins (Lloyd et al., 2001b). To avoid this potential pitfall, we examined the effect of changes in the coding sequence of the InvJ secretion signal in the context of the full-length InvJ protein. We introduced +1 or −1 reading frameshift mutations downstream of the ATG initiation codon of InvJ and corrected the frameshift either 8 or 16 codons downstream. Consistent with what was observed in the InvJ–PhoA chimeric constructs, the frameshifts did not affect the secretion of InvJ (Fig. 5). Taken together, these results indicate that the InvJ secretion signal can tolerate significant changes in its primary amino acid sequence or that the information for secretion resides in the mRNA.

Figure 5.

Introduction of frameshift mutations in the secretion signal of InvJ does not alter its secretion. Frameshift mutations (+1 and −1) were introduced immediately after the initiation codon of InvJ, and the shift was corrected either 8 or 16 residues after. The resulting constructs were expressed in an invJ mutant strain of S. typhimurium, and whole-cell lysates and culture supernatants of the resulting strains were analysed for the presence of InvJ by Western immunoblot as described in Experimental procedures.

Mutations in the mRNA coding sequence do not affect InvJ secretion

The observation that the InvJ secretion signal could tolerate significant changes in its primary amino acid sequence prompted us to examine the possibility that the InvJ secretion signal could be encoded by its mRNA. For this purpose, we introduced changes in the invJ codons 4–7, which are critical for secretion. Several constructs were generated with different permutations at nucleotide positions 12, 13, 14, 15, 18 and 21, corresponding to codons 4, 5, 6 and 7. The introduced mutations resulted in drastic alterations in the mRNA sequence and its predicted secondary structure (as predicted by mfold; data not shown) without affecting the actual coding sequence. The resulting constructs were introduced into an S. typhimurium invJ mutant and tested for their expression and secretion. The invJ mRNA sequence tolerated significant changes without affecting its secretion and/or expression (Fig. 6). Even the simultaneous introduction of changes in all four of the invJ codons required for secretion did not impair the secretion and/or expression of InvJ (Fig. 6). We also tested the effect of changes in the mRNA sequence of codons 4–7 on the expression and secretion of an InvJ1−7–PhoA chimeric protein. Consistent with the results obtained with wild-type InvJ, the simultaneous introduction of changes in these codons did not affect the relative secretion of the reporter protein, although it slightly impaired the efficient expression of this construct (Fig. 7). Taken together, these results argue against the involvement of the mRNA sequence of the InvJ signal sequence in its secretion.

Figure 6.

Effect of mutations in the mRNA of the secretion signal on InvJ secretion. Mutations that affect the mRNA but not the coding sequence of the secretion signal were introduced into the full-length invJ sequence, and the resulting constructs were expressed in an S. typhimurium invJ strain. The levels of InvJ in whole-cell lysates and culture supernatants were determined by Western immunoblot as described in Experimental procedures.

Figure 7.

Effect of mutations in the mRNA of the secretion signal on InvJ secretion and its ability to direct the secretion of alkaline phosphatase. Mutations that affect the mRNA but not the coding sequence of the secretion signal were introduced into an InvJ1−7–PhoA construct, and the resulting constructs were expressed in wild-type S. typhimurium. The levels of PhoA in whole-cell lysates and culture supernatants were determined by Western immunoblot as described in Experimental procedures.

Sequences 5′ of invJ are required for its efficient expression but not for its secretion

Although the mutational analysis described above does not support the involvement in InvJ secretion of the mRNA sequence encoding the InvJ amino acid secretion signal sequence, it remained possible that sequences 5′ of invJ located within the invI open reading frame (ORF) could contribute to secretion as both invI and invJ are in the same transcriptional unit. Indeed, the start of the invJ ORF overlaps the stop codon of invI. To address this issue, we compared the expression and secretion of InvJ when expressed in conjunction with the upstream gene invI or when expressed from a construct (pHR22) encoding just invJ and its predicted ribosome binding site (i.e. 22 bp of upstream invI). As shown in Fig. 8, although InvJ was efficiently expressed and secreted when expressed downstream from invI, it was poorly expressed from an equivalent construct containing just its predicted ribosome binding site. Consistent with its poor expression, this construct was unable to complement an invJ mutant (Fig. 8). These results suggested that the cys expression of invI is required for invJ expression and/or that there are sequences within the invI ORF that are required for invJ expression. To distinguish between these possibilities, we constructed a plasmid encoding invJ plus 47 nucleotides from the 3′ end of the invI ORF. This plasmid expressed wild-type levels of InvJ, which was secreted efficiently. These results indicate that: (i) the cys expression of InvI is not required for InvJ expression and/or secretion; and (ii) there are sequences within the invI ORF that are required for its efficient translation. However, owing to the lack of expression of invJ in the absence of upstream sequences, we could not address the potential requirement for these sequences in the secretion of InvJ. To address this issue, we constructed a plasmid (pHR475) in which the translational control sequences of invJ were replaced by heterologous translational sequences derived from the pBAD24 expression vector (Guzman et al., 1995). In this case, InvJ was efficiently expressed and secreted, indicating that no mRNA sequences 5′ from the invJ ORF are required for its secretion. Taken together, these results indicate that, although there are sequences located upstream of invJ that are required for its efficient expression, those sequences are not essential for InvJ secretion.

Figure 8.

Sequences 5′ of invJ are required for its efficient expression but not for its secretion. Plasmids encoding full-length invI and invJ (pHR104), 47 nucleotides of the 3′ end of invI and full-length invJ (pSB1631), 22 nucleotides of the 3′ end of invI and full-length invJ (pHR22) or invJ fused to translational control signals from the expression vector pBAD24 (pHR475) were introduced into an invJ S. typhimurium mutant strain (SB302). The levels of InvJ in whole-cell lysates and culture supernatants of wild type (SB300) and the mutant strains were determined by Western immunoblot as described in Experimental procedures.

Functional delineation of the InvJ secretion signal

We took advantage of the strong phenotypes associated with InvJ to confirm the delineation of the type III secretion signal of InvJ using several very robust functional assays. These assays have the advantage that they do not rely on the measurement of the actual secretion of InvJ and therefore provide us with an independent means of assessing the accuracy of the delineation of the InvJ secretion signal carried out by biochemical methods. We first examined different invJ mutant constructs for their ability to control the length of the needle substructure of the needle complex. We found a perfect correlation between the ability of InvJ to be secreted and its ability to control the length of the needle substructure of the needle complex (Fig. 9). All frameshifted mutants also behaved in a manner indistinguishable from the wild-type protein. We also tested the secretion signal mutants of InvJ for their ability to restore invasion to a strain carrying a loss-of-function invJ mutant. Similar to the needle length control functions of InvJ, we found that mutations that affect the secretion of InvJ were unable to restore the invasiveness of an invJ-deficient strain of S. typhimurium. These results are summarized in Table 1. Taken together, these studies provide a functional corroboration of the delineation of the InvJ secretion signal sequence and confirm that InvJ secretion is required for its function.

Figure 9.

Complementation of the needle length control function of InvJ by plasmids encoding different mutations in the secretion signal. Electron micrographs of negatively stained untreated (top) or osmotically shocked (bottom) preparations of S. typhimurium invJ mutants carrying the indicated plasmids. The characteristics of the mutants encoded in the different plasmids are described in Table 1 (bar = 1 µm, top, and 100 nm, bottom).

Table 1. . Summary of the phenotypes of the different InvJ mutant constructs.
PlasmidProtein% InvasionaSecretionbNeedlec
  • a

    . Values are the percentage of the initial inoculum that survived the gentamicin treatment and represent the mean ± standard deviation of three independent determinations. The invasion values of the wild-type strain SB300 and the invJ mutant SB302 were 49 ± 0.61 and 0.23 ± 0.03 respectively. All data were obtained using the SB302 mutant strain.

  • b

    . Secretion was evaluated by Western blot analysis.

  • c

    . The needle component of the needle complex was examined by electron microscopy.

pHR104InvJ1−336  37 ± 2.1+Normal
pHR485InvJ1−179–PhoA35−472ND+ND
pHR778InvJ1−52–PhoA35−472ND+ND
pHR785InvJ1−25–PhoA35−472ND+ND
pHR787InvJ1−15–PhoA35−472ND+ND
pHR789InvJ1−11–PhoA35−472ND+ND
pHR791InvJ1−8–PhoA35−472ND+ND
pHR6InvJ1−7–PhoA35−472ND+ND
pHR4InvJ1−6–PhoA35−472NDND
pHR2InvJ1−5–PhoA35−472NDND
pHR99InvJ3−7–PhoA35−472ND+ND
pHR100InvJ4−7–PhoA35−472ND+ND
pHR101InvJ5−7–PhoA35−472NDND
pHR102InvJ6−7–PhoA35−472NDND
pHR71InvJ1−336  36 ± 2.3+Normal
pHR72InvJ(Δ2)−336  35 ± 1.1+Normal
pHR73InvJ(Δ2−3)−336  36 ± 1.7+Normal
pHR74InvJ(Δ2−4)−3360.88 ± 0.1Long
pHR82InvJ(Δ2−5)−3360.72 ± 0.07Long
pHR83InvJ(Δ2−6)−3360.41 ± 0.1Long
pHR84InvJ(Δ2−7)−3360.58 ± 0.08Long
pHR85InvJ(Δ2−8)−3360.34 ± 0.07Long
pHR9InvJ1−15(+1)–PhoA35−472ND+ND
pHR11InvJ1−15(−1)–PhoA35−472ND+ND
pHR63InvJ1−7(−1)–PhoA35−472ND+ND
pHR61InvJ1−6(−1)–PhoA35−472NDND
pHR58InvJ1−15(+1)16−336  38 ± 1.7+Normal
pHR59InvJ1−15(−1)16−336  35 ± 1.3+Normal
pHR76InvJ1−7(−1)8−336  36 ± 1.9+ND
pHR105InvJ1−6(−1)8−3360.78 ± 0.11ND
pHR385InvJ1–7 (4,5,6,7 = GTAAGCGCCGTA)  36 ± 1.4+Normal
pHR386InvJ1–7 (4,5,6,7 = GTTAGTGCGGTT)  36 ± 1.9+Normal
pHR448InvJ1–7 (4,5,6,7 = GTGTCAGCCGTA)  35 ± 1.1+ND
pHR452InvJ1–7 (4,5,6,7 = GTAAGCGCCGTC)  34 ± 2.3+ND
pHR453InvJ1–7 (4,5,6,7 = GTGAGCGCCGTA)  35 ± 1.8+ND
pHR454InvJ1–7 (4,5,6,7 = GTATCAGCCGTA)  36 ± 1.0+ND
pHR455InvJ1–7 (4,5,6,7 = GTAAGCGCTGTA)  36 ± 1.6+ND
pHR459InvJ1–7 (4,5,6,7 = GTGTCAGCCGTA)–PhoA35−472ND+ND
pHR458InvJ1–7-(4,5,6,7 = GTAAGCGCCGTA)–PhoA35−472ND+ND
pHR22 invJ+22 bp of upstream sequence0.56 ± 0.09ND
pSB1631 invJ+47 bp of upstream sequenceND+ND
pHR475 invJ+heterologous translational sequencesND+ND

Discussion

The mechanisms by which proteins destined to travel through the TTSS are recognized by the secretion machinery are poorly understood. We sought to define the signal(s) that direct S. typhimurium InvJ protein to the secretion apparatus. Using reporter fusion proteins as well as sequential deletion mutagenesis of the full-length InvJ protein, we found that the secretion signal of InvJ is encoded within its first seven amino acids and, in particular, within a region encompassing residues 4–7. The secretion signal was found to be necessary and sufficient to mediate InvJ secretion because: (i) it was able to mediate the secretion of a PhoA reporter protein; and (ii) its removal resulted in an InvJ protein that, although expressed at wild-type levels, was not secreted and was unable to complement an invJ-null mutant strain.

We sought to distinguish between the requirement for a specific mRNA or amino acid sequence to direct the secretion of InvJ by constructing a number of frameshifted mutant derivatives. Introduction of +1 or −1 reading frame shifts in the first 15 or seven codons of InvJ or the InvJ–PhoA chimeric protein did not alter their secretion. In addition, the InvJ frameshifted mutants complemented an invJ-null mutant strain of Salmonella in various sensitive functional assays in a manner indistinguishable from that of wild type. As the frameshifts resulted in sequences that encode different amino acid residues with completely different biochemical properties (i.e. charge, side-chains, etc.), these results would support the notion that secretion depends on the mRNA rather than on the actual amino acid sequence. However, introduction of mutations that drastically altered the mRNA sequence and its predicted structure without altering the coding sequence had no effect on the secretion and/or expression of InvJ, arguing that the mRNA encoding the signal sequence is not the actual determinant of secretion.

These results present an apparent conundrum: how could a polypeptide sequence contribute to specific secretion when it is so tolerant to changes in its amino acid sequence composition? This conundrum is particularly pertinent in the case of InvJ, as this protein is unlikely to posses a cognate chaperone. Although the possibility that the upstream gene invI may encode for a putative InvJ chaperone has been raised (Collazo et al., 1995), a loss-of-function mutation in invI does not affect the expression and/or stability of InvJ (Collazo et al., 1995), and studies have failed to demonstrate a physical interaction between InvI and InvJ (M. Zierler and J. E. Galán, unpublished results). Furthermore, unlike type III secreted proteins that posses a cognate chaperone, InvJ is not secreted in the absence of its amino-terminal secretion signal. This also argues against the possibility of the existence of an additional chaperone-mediated pathway that could compensate for the absence of or mutations in its secretion signal. Given the constraints imposed by the estimated diameter of the translocation channel of the needle complex (≈ 30 Å), it is likely that InvJ is secreted and recognized by the secretion apparatus in an unfolded manner. In this context, it is possible that a variety of polypeptide sequences (although certainly not all) may be suitable for recognition by the secretion apparatus provided that they do not rapidly acquire secondary or tertiary structural features that would be incompatible with substrate recognition. A similar model has been proposed for the mechanisms of recognition of flagellar proteins by the flagellar export apparatus (Karlinsey et al., 2000). This model would be in keeping with the observation that drastically different secretion signals, resulting from the introduction of frameshift mutations, are fully competent to target InvJ for secretion. It would also be consistent with the failure to identify robust consensus sequences by comparing the secretion signals of different type III secreted proteins despite the observation that the secretion apparatuses are capable of engaging heterologous substrates. Indeed, mechanisms of recognition of unfolded polypeptides of vastly different amino acid sequences are well described in the case of bacterial chaperones such as SecB or members of the GroEL/GroES family (Driessen et al., 2001; Hartl and Hayer-Hartl, 2002). It is therefore possible that functionally equivalent mechanisms may be used by proteins associated with the type III secretion apparatus to recognize polypeptides destined to be secreted through this pathway. However, type III secretion is very specific, and only a very small number of proteins are secreted via this pathway. It is unknown whether amino-terminal sequences that could potentially fulfil the proposed requirements to function as type III secretion signals occur in proteins that are not secreted by the type III pathway. If this is the case, other determinants of specificity must operate to ensure the specific targeting of type III secreted proteins that lack cognate chaperones. An attractive idea is the possibility that such determinants may be encoded within the mRNA (Anderson and Schneewind, 1997; 1999). However, in the case of InvJ, we found no evidence to support this hypothesis. Secretion determinants are unlikely to be located within the mRNA that encodes the actual signal sequence, as introduction of mutations that drastically altered this mRNA sequence and its predicted folded structure had no effect on secretion. We identified determinants involved in the regulation of invJ translation located within 47 bp of the 3′ end of the upstream invI ORF. However, those determinants do not appear to be essential for secretion, as placing invJ under the control of a heterologous translational control sequence did not impair InvJ secretion. Although it remains a possibility that the efficient translation of invJ mediated by the heterologous signals may somehow relieve the need for mRNA targeting sequences, our data indicate that efficient secretion of InvJ can take place in the apparent absence of mRNA regulatory signals. More studies will be required not only to validate this model of InvJ recognition by the secretion machinery but also to examine its potential generalization to other type III secreted proteins.

Experimental procedures

Bacterial strains and growth conditions

The wild-type S. typhimurium strain SL1344 and its mutant derivatives carrying loss-of-function mutations in invJ (Collazo et al., 1995) or invA (Galán et al., 1992) have been described previously. Strains were grown on L-agar or in L-broth supplemented with 0.3 M sodium chloride to allow optimal expression of the components of the invasion-associated type III secretion system. When required, the following antibiotics were added at the concentrations indicated: kanamycin, 50 µg ml−1; ampicillin, 100 µg ml−1; streptomycin, 100 µg ml−1; tetracycline, 10 µg ml−1.

Plasmid constructions

All plasmid used in this study are listed in Table 1. To construct the different InvJ–PhoA fusions, we made use of plasmid pSK2B (a kind gift from C. Roy, Yale University, New Haven, CT, USA), which carries a truncated phoA gene of E. coli K-12 coding for amino acids 35–472 of the alkaline phosphatase (PhoA). Expression of InvJ is translationally coupled to the expression of the upstream gene invI. Therefore, to ensure appropriate expression of InvJ, unless stated otherwise, the sequence of invI in identical context to that of wild type was included in all constructs used in these studies. Different fragments of invJ and the upstream gene invI were amplified by polymerase chain reaction (PCR) and cloned into the SacI and BamHI restriction sites of pSK2B resulting in different InvJ–PhoA fusion proteins. All constructs were subsequently subcloned into the low-copy plasmid vector pWSK29 using the SacI and KpnI restriction enzymes, so that expression was driven by the lac promoter. This general strategy was used to construct the following plasmids: pSB485 (InvJ1−179–PhoA), pSB778 (InvJ1−52–PhoA), pSB785 (InvJ1−25–PhoA), pSB787 (InvJ1−15–PhoA), pSB789 (InvJ1−11–PhoA), pSB791 (InvJ1−8–PhoA), pHR6 (InvJ1−7–PhoA), pHR4 (InvJ1−6–PhoA), pHR2 (InvJ1−5–PhoA).

To construct another set of chimeric InvJ–PhoA proteins, we progressively introduced single amino acid deletions at the N-terminus of the first seven amino acids of InvJ. An upstream PCR fragment containing invI and the invJ initiation codon and a downstream PCR fragment harbouring the genetic information for the respective deletion mutant of InvJ were assembled as cassettes with the sites SacI–BamHI–KpnI and inserted between the SacI–KpnI sites of pWSK29 to yield pHR99 (InvJ3−7–PhoA), pHR100 (InvJ4−7–PhoA), pHR101 (InvJ5−7–PhoA) and pHR102 (InvJ6−7–PhoA).

Further hybrid InvJ–PhoA proteins were constructed containing +1 or −1 reading frameshifts by either adding or deleting a guanine nucleotide immediately downstream of the ATG start codon of InvJ. The different frameshifts were introduced into 3′ PCR primers, and amplified DNA fragments containing the genetic information for InvI and frameshifted InvJ were cloned into the SacI–BamHI restriction sites of pSK2B. Subcloning of the resulting invJ–phoA gene fusions into pWSK29 was performed as described above, resulting in pHR9 (InvJ1−15(+1)–PhoA), pHR11 (InvJ1−15(−1)–PhoA), pHR63 (InvJ1−7(−1)–PhoA) and pHR61 (InvJ1−6(−1)–PhoA).

To construct plasmids expressing different InvJ mutants, the wild-type invI and invJ genes were amplified and cloned into the SacI–EcoRI sites of pWSK29 resulting in pHR104. In this plasmid (and its derivatives), expression is driven by the lac promoter. A series of N-terminal deletions of InvJ were constructed as follows. An upstream PCR fragment containing invI and the invJ initiation codon and a downstream PCR fragment harbouring the genetic information for the respective N-terminal deletion of InvJ were assembled as cassettes with the sites SacI–BamHI–EcoRI and inserted between the SacI–EcoRI sites of pWSK29. The BamHI restriction site was introduced in the coding sequence of InvJ immediately downstream of the ATG start codon, yielding plasmids pHR71 (InvJ1−336), pHR72 (InvJ[Δ2]−336), pHR73 (InvJ[Δ2−3]−336), pHR74 (InvJ[Δ2−4]−336), pHR82 (InvJ[Δ2−5]−336), pHR83 (InvJ[Δ2−6]−336), pHR84 (InvJ[Δ2−7]−336) and pHR85 (InvJ[Δ2−8]−336). Introduction of this restriction site resulted in the addition of a glycine (G) and serine (S) immediately downstream of the start codon, which did not affect either the expression or the secretion of these constructs.

Various full-length InvJ proteins containing +1 or −1 reading frameshifts downstream of the ATG start codon of InvJ were constructed using the plasmids pHR9, pHR11, pHR63 and pHR61 coding for InvJ–PhoA chimeric proteins with the respective frameshifts. Briefly, these plasmids were digested with BamHI–KpnI to cut out the truncated phoA gene and ligated to BamHI–KpnI DNA fragments containing InvJ7−336, InvJ8−336 or InvJ16−336. The resulting plasmids, pHR58, pHR59, pHR76 and pHR105, encode InvJ1−15(+1)16−336, InvJ1−15(−1)16−336, InvJ1−7(−1)8−336 and InvJ1−6(−1)7−336 respectively.

Plasmids encoding invJ with changes in codons 4, 5, 6 and 7, resulting in alterations in the mRNA sequence without affecting the amino acid coding sequence, were constructed following a strategy similar to the one outlined above. Briefly, an upstream PCR fragment containing invI and the initiation codon of invJ and a downstream PCR fragment harbouring different permutations at nucleotide positions 12–21 were assembled as cassettes with the sites SacI–BamHI–EcoRI and inserted between the SacI–EcoRI sites of pWSK29. Plasmid vectors pHR385 and pHR386 carry changes in invJ codons 4–7, pHR452 in codons 4–6, pHR453 in codons 5–7, pHR454 in codons 4, 6 and 7, pHR455 in codons 4, 5 and 7, and pHR448 in codons 6 and 7. The construction of InvJ–PhoA chimeric proteins with changes in the invJ codons 4, 5, 6 and 7, resulting in alterations in the mRNA sequence without affecting the amino acid coding sequence, was carried out by PCR following a similar strategy to the one outlined above. Plasmid pHR22 carries an amplified PCR fragment containing the invJ ORF plus 22 bp of the upstream invI sequence cloned into the SacI–EcoRI sites of the vector pWSK29 (Wang and Kushner, 1991). Plasmid pHR475 encodes the invJ ORF (obtained by PCR) cloned into the NcoI–SmaI sites of the expression vector pBAD24 so that expression of invJ is under the control of heterologous translational and transcriptional sequences provided by the vector (Guzman et al., 1995). Plasmid pSB1631 carries invJ plus 47 bp of upstream sequence. This plasmid was constructed by cloning the KpnI–BamHI fragment of pSB475 (Collazo et al., 1995) into the expression vector pWSK29 (Wang and Kushner, 1991). All constructs were verified by DNA sequencing.

Quantification of bacterial entry

The ability of the different bacterial strains to enter into a culture of intestinal Henle-407 cells was evaluated as described previously (Galán and Curtiss, 1989).

Preparation and analysis of culture supernatant proteins

Culture supernatant proteins were prepared as follows. Briefly, 10 ml bacterial supernatants were passed through a 0.45-µm pore size syringe filter to remove bacteria. Protein in the bacteria-free medium was precipitated by the addition of cold trichloroacetic acid (TCA) to 10% (v/v) and incubated on ice for 2 h. The protein was collected by centrifugation at 4°C, 10 000 g for 20 min. Pellets were washed in 0.8 ml of cold acetone, dried and resuspended in PBS buffered with 80 mM Tris-HCl, pH 8.0. Samples corresponding to 400 µl of whole bacterial culture and 2 ml of culture supernatant were separated in a 10% SDS-polyacrylamide gel and transferred to nitrocellulose membranes. InvJ and InvJ–PhoA chimeric proteins were detected by immunoblot analysis. Western blots were treated with a monoclonal antibody against InvJ or a polyclonal antibody directed against PhoA, followed by incubation with a horseradish peroxidase-labelled anti-mouse or anti-rabbit antibody. Blots were developed using a chemiluminescense detection kit.

Electron microscopy

Osmotically shocked cells were prepared as described elsewhere (Kubori et al., 1992). Samples were negatively stained with 2% phosphotungstic acid (pH 7.0) and observed under the electron microscope (EM410, Philips). Micrographs were taken at an accelerating voltage of 80 kV. A phenotype was assigned (presence or absence of the needle substructure) to a given strain when at least 20 cells had been scored. A wild-type cell exhibits ≈ 20 needle complexes in the visible membrane edge of the osmotically shocked bacteria. Therefore, by examining 20 bacterial cells, a minimum of 400 structures were scored for the presence or absence of the needle portion.

Acknowledgements

We thank members of the Galán laboratory for critical review of this manuscript. H.R. was supported by the ‘AIDS-Stipendienprogramm’ from the Bundesministerium für Bildung, Wissenschaft, Forschung und Technologie/Germany. T.K. was supported by a fellowship from the Human Frontier Science programme. This work was supported by Public Health Service grant AI30492 from the National Institutes of Health to J.E.G.

Ancillary