Investigation of the role of the BAM complex and SurA chaperone in outer-membrane protein biogenesis and type III secretion system expression in Salmonella

In Gram-negative bacteria, the outer membrane (OM) is an asymmetrical bilayer containing phospholipids in the inner leaflet, and lipopolysaccharide in the outer leaflet. Outer-membrane proteins (OMPs) are inserted in this membrane. The assembly of OMPs in the OM requires the BAM complex and the periplasmic chaperones SurA, DegP (also known as HtrA) and Skp, which are thought to deliver most OMPs to the BAM complex. In Escherichia coli, the BAM complex is composed of the OMP BamA (YaeT) and four lipoproteins: BamB (YfgL), BamC (NlpB), BamD (YfiO) and BamE (SmpA) (Malinverni et al., 2006; Ruiz et al., 2005; Sklar et al., 2007a; Wu et al., 2005). In this complex, both BamA and BamD are essential in E. coli, and depletion of these proteins results in a severe impairment of OMP assembly (Malinverni et al., 2006; Wu et al., 2005). In contrast, BamB, BamC and BamE are not essential in E. coli. Loss of BamB causes a reduction in OMP assembly, and consequently an increased susceptibility of the bacteria to antibiotics (Charlson et al., 2006; Malinverni et al., 2006; Onufryk et al., 2005; Wu et al., 2005). In contrast, the loss of BamC or BamE results in a minor OMP targeting defect, which is greater in a bamE mutant than in a bamC mutant (Charlson et al., 2006; Malinverni et al., 2006; Onufryk et al., 2005; Wu et al., 2005). Currently, the model proposed is that OMPs are primarily delivered to the BAM complex by the periplasmic chaperone SurA under normal growth conditions, while DegP and Skp play a role mostly under stress conditions (Sklar et al., 2007b).

In the enteropathogenic bacterium Salmonella enterica subsp. enterica serovar Enteritidis, commonly named S. Enteritidis, BamB has recently been shown to have a role in virulence, and particularly in the expression of type III secretion systems (T3SSs), which are major virulence factors (Amy et al., 2004; Fardini et al., 2007). Loss of BamB results in a virulence defect in vivo in a murine model of systemic infection, and in a reduced colonization of caeca and spleens in 1-day-old chicks. In vitro, the ability of a bamB mutant to invade epithelial cells, and to secrete T3SS-1 effectors and flagellar proteins, is impaired (Amy et al., 2004; Fardini et al., 2007). These in vivo and in vitro phenotypes are related to the transcriptional downregulation of SPI-1, SPI-2 and flagella genes involved in the biosynthesis of the T3SS-1, T3SS-2 and flagella, respectively (Fardini et al., 2007). T3SSs play a crucial role in host infection by Salmonella. They allow the bacteria to inject effector proteins into the host cytosol, and these proteins hijack the eukaryotic cellular machinery for the benefit of the bacteria. T3SS-1 is mainly responsible for bacterial invasion into eukaryotic cells (Schlumberger & Hardt, 2005). T3SS-2 plays an important role in intracellular survival, and replication of Salmonella in the vacuole of the host cells, and is therefore essential for systemic dissemination of the bacteria (Abrahams & Hensel, 2006). The flagella share a similar structural design with T3SS (Macnab, 2004), and confer motility to Salmonella, and thus contribute to bacterial interaction with the intestinal epithelium (van Asten et al., 2004). Salmonellae are not the only bacteria in which a role for BamB in virulence has been demonstrated; a defect in invasion ability has been observed in a bamB mutant of the adherent invasive E. coli LF82 strain (Rolhion et al., 2005).

We have previously shown that the role of BamB in OMP targeting to the OM is conserved in S. Enteritidis (Fardini et al., 2007). Therefore, BamB is directly or indirectly involved in two distinct pathways: OMP biogenesis and the control of T3SS expression. A similar reasoning could be proposed for SurA, since surA mutants have an OMP assembly defect in E. coli, and have been shown to be impaired in their ability to invade eukaryotic cells; in Salmonella this property is mainly related to T3SS-1 (Humphreys et al., 2003; Rouviere & Gross, 1996; Sydenham et al., 2000). However, no data are available concerning a possible link between OMP biogenesis and T3SS expression. To answer this question, we studied the role of the different proteins of the BAM complex, and of the SurA and DegP chaperones, in these two phenotypes in Salmonella.

Bacterial strains and culture conditions.
Bacterial strains used in this work are listed in Table 1. The S. Enteritidis LA5 strain is a field isolate from infected chicken, and it is resistant to nalidixic acid (20 µg ml^–1) (Allen-Vercoe et al., 1997). Bacteria were routinely grown in Luria–Bertani (LB) broth at 37 °C overnight, with shaking. Antibiotics were added to cultures at the following concentrations: spectinomycin (Sp), 100 µg ml^–1; nalidixic acid (Nx), 20 µg ml^–1; kanamycin (Km), 50 µg ml^–1; chloramphenicol (Cm), 30 µg ml^–1; and carbenicillin (Cb), 100 µg ml^–1. When necessary, bacteria were cultured in LB broth supplemented with 0.3 M NaCl (Arricau et al., 1998). For growth analysis, 150 ml fresh LB medium was inoculated with an overnight culture to a final concentration of about 10⁶ bacteria ml^–1, and cultured for 24 h at 37 °C, with shaking. Growth was recorded by measurement of the OD at 600 nm, and by plating serial dilutions of the culture on tryptone soy agar (TSA).

Table 1. Strains and plasmids used in this study

Antibiotic sensitivity.
The sensitivity of the different strains to rifampicin (Rf), erythromycin (Em), vancomycin (Vm) and bacitracin (Bc) was assessed by a disk diffusion assay using 6 mm filter paper disks (Bio-Rad), as previously described (Fardini et al., 2007). At least three independent experiments were carried out for each strain.

Mutagenesis.
Mutants were obtained using the λ-Red mutagenesis system, modified as described (Figueroa-Bossi et al., 2006). The sequences of the primers used are listed in Table 2. The construction of the orf mutant was carried out as follows. The antibiotic resistance cassettes Cm^R and Km^R, of pKD3 and pKD4, respectively, were amplified with primers ORF-P1 and ORF-P2, and the PCR product was electroporated in the S. Enteritidis LK5hsdR strain harbouring pKD46, using a Micropulser (Bio-Rad) in accordance with the manufacturer's recommendations. Cm^R or Km^R Cb^S clones were selected, and the insertion was controlled by PCR with primers ORF1/ORF2. The mutation was then transduced in the S. Enteritidis LA5 strain using phage P22HT105int. After PCR verification of the transductants with the primers described above, plasmid pCP20 was introduced into the orf mutant of strain LA5 by electroporation to remove the Cm^R or Km^R cassette. Transformants were first selected at 30 °C on the basis of their resistance to Cb, then cultured at 42 °C without antibiotics to eliminate the plasmid pCP20, and finally tested for the loss of resistance to Cb and Cm/Km. The scar sequences, and the deletion limits of the Δorf mutants, were checked by sequencing.

Table 2. Primers used in this study

Construction of the plasmid pACbamD.
The complete coding sequence of bamD, from the chromosomal DNA of strain LA5, was amplified by PCR with primers yfiO3/yfiO4 (Table 2). After restriction with SmaI and HindIII, the 1 kb PCR product was cloned into the SmaI and HindIII sites of pACYC177, to generate the pACbamD plasmid. The transcription of bamD in this recombinant plasmid was under the control of the promoter of the Km^R gene of the pACYC177 vector. The sequence of the pACbamD insert was checked by sequencing.

Protein analysis.
In order to study membrane proteins, bacteria were grown in LB broth containing 0.3 M NaCl until the OD at 600 nm reached 0.8–1.0. Total membrane proteins were extracted as described previously (Fardini et al., 2007). The proteins were analysed by electrophoresis in a 12 % (w/v) SDS-polyacrylamide gel, and either stained with Coomassie Brilliant Blue G-250 (Neuhoff et al., 1988), or transferred onto a nitrocellulose membrane (Protran; Schleicher and Schuell). Immunoblotting was performed with polyclonal rabbit antisera raised against OmpA (1 : 2000); the antisera were kindly provided by R. Lloubes (CNRS, Marseilles, France). Blots were revealed using horseradish-peroxidase-labelled goat anti-rabbit antibodies (1 : 5000; Dako) and the Amersham ECL plus Western blotting detection system (GE Healthcare). Identification of the proteins on the Coomassie-Brilliant-Blue-stained gels was performed by comparison of the profiles with profiles obtained previously for the wild-type (WT) strain or strains carrying mutations in ompC, ompF, ompD or ompA genes (unpublished results), and/or by N-terminal microsequencing. At least three independent experiments were carried out for each strain.

Secreted proteins were obtained from bacteria cultured in LB broth containing 0.3 M NaCl until the OD at 600 nm reached 1.8–2.0, as described previously (Amy et al., 2004). Proteins were separated by electrophoresis in a 10 % (w/v) SDS-PAGE, and they were either stained with Coomassie Brilliant Blue G-250, or transferred onto a nitrocellulose membrane and immunoblotted with polyclonal rabbit antisera raised against SipA (1 : 1000) or H:g,m flagellin (1 : 500), as previously described (Amy et al., 2004). At least three independent experiments were carried out for each strain.

RNA extraction and real-time RT-PCR.
RNA was recovered from bacteria cultured in LB broth containing 0.3 M NaCl. Total bacterial RNA was obtained as described (Glatron & Rapoport, 1972). Potential DNA contamination was eliminated by DNase I treatment (Invitrogen) in accordance with the manufacturer's instructions. Absence of DNA was confirmed by amplification of the tufA gene by PCR, as previously described (Fardini et al., 2007). The amount, purity and quality of the RNA were evaluated by OD measurement at 230, 260 and 280 nm, and by gel electrophoresis followed by ethidium bromide staining. A 1 µg quantity of RNA was then reverse transcribed by the avian myeloblastosis virus reverse transcriptase (Promega), with random hexamer primers (Promega), in accordance with the manufacturer's protocol. Real-time PCR (LightCycler; Roche) was carried out to quantify the transcriptional levels of sipA, hilA and fliD, and the housekeeping gene tufA (McIngvale et al., 2002). Primers and conditions used to perform these real-time PCRs have been described previously (Fardini et al., 2007). The relative number of cDNA genes was quantified in duplicate from three independent RNA extractions and bacterial cultures.

Invasion assay.
Bacteria were grown in LB broth overnight without shaking, or for 6 h with shaking. The human adenocarcinoma cell line HT-29 (European Collection of Animal Cell Cultures no. 85061109) (Fogh & Trempe, 1975) was used between passages 27 and 67. Cell monolayers were grown to confluence in 24-well plates, as previously described (Roche et al., 2003). Monolayers composed of between 1.3x10⁶ and 1.5x10⁶ cells per well were then infected with bacteria at a m.o.i. of 30 for 2 h at 37 °C. Bacterial invasion was then quantified by the gentamicin protection assay, as described previously (Amy et al., 2004). The rate of invasion was expressed by dividing the number of gentamicin-resistant bacteria by the number of bacteria deposited per well, and then multiplying by 100. Three independent experiments were performed in duplicate. Results were then compared using a Kruskal–Wallis non-parametric test, and analysed using Dunn's multiple comparison test, using Prism software (GraphPad).

BamD is not essential in S. Enteritidis
In order to study the role of the different proteins of the BAM complex in S. Enteritidis, isogenic deletion mutants of the different genes encoding the proteins of the complex were constructed. All our attempts to construct a bamA-deletion mutant failed, suggesting that bamA is essential in S. Enteritidis, as it is in E. coli and Neisseria meningitidis (Voulhoux et al., 2003; Wu et al., 2005). In contrast, the ΔbamC, ΔbamD and ΔbamE mutants were obtained easily. ΔbamC and ΔbamE mutants had growth rates similar to that of the WT, and the size of the colonies formed by these mutants on agar plates was equivalent to that of the WT (data not shown). In contrast, several growth-related defects were observed for the ΔbamD mutant. Colonies formed by this mutant were smaller than those of the WT. Moreover, in LB medium, the ΔbamD mutant had a growth rate similar to that of the WT at the beginning of the exponential growth phase, but grew slower in the late-exponential phase and reached the stationary phase much earlier (Fig. 1). After 24 h of culture, the ΔbamD mutant showed a decreased number of viable cultivable cells (Fig. 1). Similar results were obtained in LB medium containing 0.3 M NaCl, and in M9 minimal medium (data not shown). Subcultures of this mutant resulted in a heterogeneous population (small and normal-sized colonies) (data not shown), possibly suggesting the occurrence of suppressive mutations. All these characteristics of the ΔbamD mutant indicate an important role for BamD in Salmonella growth, although, unlike in E. coli, it is not essential in Salmonella (Malinverni et al., 2006; Onufryk et al., 2005).

(16K): Fig. 1. BamD is important but not essential for S. Enteritidis growth in vitro. The growth curves for the S. Enteritidis WT strain LA5 (•), the ΔbamD mutant (), the bamDΔ227 mutant () and the bamDΔ172 mutant () are shown. Bacteria were cultured for 24 h in LB broth, with shaking. Growth was recorded by plating serial dilutions of the culture on TSA agar at regular times. Representative results from three independent cultures are shown.

Following these results, for further studies, we chose to construct strains with a BamD partial-loss-of-function, exhibiting culture-related characteristics similar to that of the WT. We constructed two different mutants: one with a mutation that has been described in E. coli, in which the last 18 codons of the BamD protein were deleted, and for which a partial-loss-of-function has been demonstrated (bamD_Δ227 mutant) (Wu et al., 2005); and one in which the BamD protein was truncated after the amino acid in position 172 (bamD_Δ172 mutant). Each of the bacterial populations of these two bamD mutants was as homogeneous as that of the WT (data not shown). Furthermore, these mutants grew as well as the WT in LB medium, except that a slightly lower number of c.f.u. of the bamD_Δ172 strain was obtained after 24 h of culture (Fig. 1). Complementation with plasmid pACbamD, which harbours the bamD gene, restored the ability of the bamD_Δ172 and ΔbamD mutants to grow like their parental strain (data not shown).

Role of the BAM complex in OMP biogenesis: involvement of BamB, BamD and, to a lesser extent, BamE in S. Enteritidis
In E. coli, the BamC, BamD and BamE lipoproteins are differentially involved in OM biogenesis (Malinverni et al., 2006; Onufryk et al., 2005; Sklar et al., 2007a). As the role of these proteins in S. Enteritidis was unknown, their involvement in OM biogenesis was assessed by comparing the membrane protein profiles of the corresponding mutants with those of the WT, and a mutant lacking BamB, a protein known to play a role in OM biogenesis in S. Enteritidis (Fardini et al., 2007). We observed that the OMP defect in the S. Enteritidis ΔbamB mutant was the same when bacteria were cultured in LB or in LB containing 0.3 M NaCl (unpublished results); therefore, using SDS-PAGE, we studied the OMP profiles of the different mutants after culture in LB containing 0.3 M NaCl, because this culture condition was then used for further analysis of T3SS production. We observed no reduction in the levels of the major OMPs (OmpA, OmpC/F and OmpD) in the ΔbamC mutant (Fig. 2a). In the ΔbamE mutant, the OMP profile showed a slight decrease in the levels of these porins compared with the WT (Fig. 2a). In contrast, and similar to the ΔbamB mutant, the two BamD partial-loss-of-function mutants showed a marked reduction in porin levels in their membranes. Furthermore, the bamD_Δ172 mutant was more affected than either the bamD_Δ227 or the ΔbamB mutant (Fig. 2a). These results were confirmed by Western blotting using an antiserum raised against OmpA (Fig. 2b).

(77K): Fig. 2. Involvement of the different lipoproteins of the BAM complex in the OMP levels in S. Enteritidis. The OMP levels were analysed by preparation of total membrane protein extracts from the S. Enteritidis WT strain LA5 and the different mutants. Strains were grown in LB supplemented with 0.3 M NaCl. Proteins were separated by 12 % SDS-PAGE, and stained with Coomassie Brilliant Blue (a), or transferred onto a nitrocellulose membrane, and probed with polyclonal antibodies raised against OmpA (b). The positions of OmpA, OmpC/F, OmpD and FliC are indicated. Results shown are from a representative of three independent cultures and protein extractions. M, Molecular mass standards.

In E. coli and S. Enteritidis, the decrease in the OMP levels resulting from the loss of BamB is related to an increase of the OM permeability to various antibiotics (Fardini et al., 2007; Ruiz et al., 2005; Wu et al., 2005). Thus, we analysed the involvement of BamC, BamD and BamE in bacterial sensitivity to different antibiotics. The sensitivity of the ΔbamC mutant to Vm, Bc, Rf and Em was the same as the WT, whereas the ΔbamE was more sensitive than the WT to Rf only, as described for E. coli (Sklar et al., 2007a). In contrast, the two BamD partial-loss-of-function mutants were more sensitive than the WT and the ΔbamB mutant to the four antibiotics assessed (Table 3).

Table 3. Involvement of the different lipoproteins of the BAM complex in antibiotic sensitivity of S. Enteritidis

Overall, these results show that in the BAM complex, in addition to BamB, BamD and to a lesser extent BamE play a role in OM biogenesis in S. Enteritidis, while the membrane is not notably disturbed by the lack of BamC.

Influence of the BAM complex on the virulence of S. Enteritidis: requirement of BamB and BamD
In order to ascertain whether Salmonella invasion could be influenced by loss of BamD, BamC or BamE, we analysed the ability of mutants lacking these proteins to penetrate HT-29 epithelial cells. The WT and the ΔbamB mutant were used as controls. The ΔbamB mutant was able to invade HT-29 cells to a significantly lesser extent than the WT (P<0.05). No significant difference was observed between either the ΔbamE or the ΔbamC mutant and the WT, although, compared with the WT, a slightly lower invasion rate, and a slightly higher invasion rate, were observed for the ΔbamE and the ΔbamC mutants, respectively. In contrast, the capacity of both BamD partial-loss-of-function mutants to invade HT-29 cells was significantly reduced compared with that of the WT (P<0.05 for bamD_Δ227 mutant versus WT, and P<0.001 for bamD_Δ172 mutant versus WT). In the culture conditions used, i.e. preculture without shaking, the bamD_Δ227 mutant showed a 10-fold decrease in its invasion rate, and the bamD_Δ172 mutant did not enter HT-29 cells (i.e. entry was not observed at the detection threshold used in the experiment). The introduction of pACbamD restored the ability of the bamD_Δ227 mutant to invade HT-29 cells (to a level that was about 50 % of the rate shown by the WT) (Fig. 3). For reasons that are unknown, this plasmid totally failed to restore the ability of the bamD_Δ172 mutant to invade HT-29 cells when bacteria were cultured overnight without shaking (data not shown). In contrast, when bacteria were precultured for 6 h with shaking, the pACbamD plasmid significantly restored the ability of the bamD_Δ172 mutant to invade HT-29 cells (to about 45 % of the rate shown by the WT) (Fig. 3). Overall, these results show that in addition to BamB, the production of BamD, but not of BamE or BamC, is required for efficient cell invasion by Salmonella.

(21K): Fig. 3. Involvement of the different lipoproteins of the BAM complex in cell invasion of S. Enteritidis. The invasion rate of the bacteria was evaluated by using the gentamicin protection assay on human HT-29 epithelial cells. Cells were infected with the different strains pre-cultured overnight without shaking (a). The cell invasion rate of the bamDΔ172 mutant and its corresponding complemented strain were compared with the WT strain after pre-culture for 6 h with shaking (b). Results correspond to the median of at least three independent experiments carried out in duplicate. Values were compared using the Kruskall–Wallis test, and analysed by Dunn's multiple comparison test: *P<0.05 WT versus ΔbamB, WT versus bamDΔ227; ***P<0.001 WT (6 h) versus bamDΔ172. The ends of each box delineate the upper and lower quartiles and the line within the box is the median. The whiskers show the spread of the data.

bamD mutants have defective T3SS-1 and flagella expression
Since the expression of the BamD lipoprotein modified the invasion ability of S. Enteritidis, we investigated T3SS-1 and flagella expression in the bamD mutants. These studies could not be performed with the bamD deletion mutant, as this mutant was not able to reach an OD at 600 nm of between 1.8 and 2.0 after culture in LB broth containing 0.3 M NaCl; these conditions are commonly used for this type of study. Moreover, when we tried to look at the secretion of T3SS-related proteins in the ΔbamD mutant at OD₆₀₀ 1.0 or 1.5, we observed a protein profile that was similar to a total protein profile, and this suggested lysis of the cells in the culture (data not shown). In contrast, the two BamD partial-loss-of-function mutants, when cultured under T3SS-1 expression conditions, showed a marked defect in the secretion of proteins compared with the WT, as observed by SDS-PAGE and staining with Coomassie Brilliant Blue. Using profiles obtained in earlier studies, and previous N-terminal sequencing and Western-blotting results obtained for a bamB mutant (Amy et al., 2004), we identified that the major proteins whose secretion was greatly affected in the bamD mutants were the T3SS-1 effector proteins SipA and SipC, and the flagella proteins FliC, FliD and FlgL (Fig. 4a). The introduction of pACbamD in these mutants restored their secretion ability, but the complementation was only partial for the bamD_Δ172 mutant, particularly for SipA secretion (Fig. 4a). Moreover, in accordance with the results of invasion tests, the ΔbamC and the ΔbamE mutants secreted these proteins at levels that were similar to those of the WT (Fig. 4a). These results were confirmed by Western blotting with sera raised against SipA and H:g,m flagellin (Fig. 4b and c).

(73K): Fig. 4. Involvement of the different lipoproteins of the BAM complex in the expression of T3SS-1 and flagella in S. Enteritidis. (a, b, c) Proteins secreted by the different strains after growth in LB supplemented with 0.3 M NaCl, with shaking. Proteins from culture supernatants were precipitated with TCA, separated by 10 % SDS-PAGE, and stained with Coomassie Brilliant Blue (a), or transferred onto a nitrocellulose membrane and probed with polyclonal antibodies raised against either SipA (b) or H:g,m flagellin (c), and revealed by chemiluminescence. (d) Transcription of genes involved in invasion and motility in strains cultured in the conditions described above. The mRNA expression of sipA, hilA and fliD was assessed by real-time RT-PCR. The expression levels of the genes were normalized to the expression level of the housekeeping gene tufA. The results correspond to the means±SEM of at least three independent cultures and RNA extractions performed in duplicate. M, Molecular mass standards.

We have previously determined that the secretion defect of T3SS-1 and flagella proteins in a ΔbamB mutant is related to a transcriptional downregulation of the genes involved in the biosynthesis of these T3SSs (Fardini et al., 2007). As no intracellular accumulation of SipA and FliD was observed in the bamD mutants when Western blotting was used (data not shown), we postulated that a mechanism similar to that observed for the ΔbamB mutant could be observed in the BamD partial-loss-of-function mutants. To address this hypothesis, transcription of the genes encoding SipA and FliD proteins was analysed using real-time RT-PCR. Interestingly, the bamD_Δ227 mutant transcribed the sipA and fliD genes at levels that were 30 and 7 times less, respectively, than those of the WT. These reductions in transcription levels were similar to those observed for the ΔbamB mutant. In the bamD_Δ172 mutant, sipA and fliD were transcribed at levels that were more than 100 times less than for the WT (Fig. 4d). In addition, the transcription of hilA, the central transcriptional activator of the T3SS-1-encoding genes, was altered in the two bamD mutants (more than 30-fold), suggesting a decrease in the transcription of all T3SS-1-encoding genes in the BamD partial-loss-of-function mutants. The introduction of pACbamD largely restored the transcription of these genes (Fig. 4d). The absence of a complete restoration of the different phenotypes tested in the bamD_Δ172 mutant, and to a lesser extent in the bamD_Δ227 mutant, could be explained by the fact that these complemented strains expressed two BamD proteins: the WT protein and a truncated BamD protein that competed with each other. Thus, when complementation in the bamD_Δ172 strain was less marked than that in the bamD_Δ227, it could have been due to the fact that the combined activity of BamD and BamD_Δ172 was lower than that of BamD and BamD_Δ227. Overall, these results demonstrate that the impaired production of T3SS-1 and flagella proteins following bamD mutations is due to a defect in the transcription of the genes encoding these systems. For the ΔbamC and ΔbamE mutants, no particular changes in the transcription of sipA, hilA and fliD were observed (Fig. 4d), and this is consistent with the efficient secretion and epithelial cell invasion observed in these mutants (Fig. 4a). Therefore, in the BAM complex, the expression of a functional BamD protein, in addition to BamB, is required for transcription of genes involved in the biosynthesis of the T3SS-1 and flagella in S. Enteritidis.

Influence of surA or degP deletion on OMP assembly and T3SS expression in S. Enteritidis
The role of the periplasmic chaperones SurA and DegP in Salmonella is not known. We therefore decided to study the role of SurA and DegP in OMP biogenesis before investigating the impact of deletion of each respective gene on T3SS expression in S. Enteritidis. When we analysed the OMP profile of the ΔdegP mutant, levels of OmpC/F, OmpD and OmpA similar to those of the WT were observed when the strains were cultured in LB containing 0.3 M NaCl (Fig. 5a, b). In contrast, the ΔsurA mutant exhibited lower levels of these major OMPs in its membrane compared with the WT (Fig. 5a, b). Interestingly, the profile obtained for the ΔsurA mutant looked qualitatively similar to that of the bamB mutant, as observed in E. coli (Ureta et al., 2007), suggesting that the porin assembly defect of these two mutants is very similar.

(74K): Fig. 5. Role of the SurA and DegP proteins in levels of OMP and expression of T3SS-1 and flagella in S. Enteritidis. The levels of OMP (a, b) and secreted proteins (c, d) were analysed as described in the legends for Fig. 2 and Fig. 4, respectively. Note that bands corresponding to the WT strain and ΔbamB mutant are combined with those of ΔsurA and ΔdegP mutants to eliminate non-relevant portions of the gel (a) or membrane (b, c, d). Results shown are from representative experiments of three independent cultures and protein extractions.

The impact of surA or degP deletion on T3SS expression was monitored by the analysis of the secretion of the flagella proteins FliC and FliD, and the T3SS-1 effector SipA, in the corresponding mutants. ΔsurA and ΔdegP mutants secreted FliC, FliD and SipA proteins at levels similar to those of the WT (Fig 5c, d). Overall, these results demonstrate that SurA is involved in OMP biogenesis in Salmonella, and that the defect of OMP assembly generated by the lack of SurA does not decrease the secretion of effectors by the T3SS-1 and flagella proteins in these bacteria. BamB has previously been shown to be involved in OMP assembly in E. coli and Salmonella (Charlson et al., 2006; Fardini et al., 2007; Onufryk et al., 2005; Ureta et al., 2007; Wu et al., 2005). In addition, we have previously demonstrated that deletion of bamB leads to a transcriptional decrease in the expression of the three T3SSs in Salmonella, and consequently to a lower ability of the organisms to interact with epithelial and macrophage cells (Amy et al., 2004; Fardini et al., 2007). These results, in addition to those showing that surA mutants of S. Typhimurium are less able than the WT to invade epithelial and macrophage cells (Sydenham et al., 2000; Humphreys et al., 2003), raised the question of a possible link between OMP biogenesis and the control of T3SS expression. To clarify this relationship, we studied the role of the BAM complex and the periplasmic chaperones SurA and DegP in S. Enteritidis.

Our study has provided what is thought to be the first description of the role of SurA, DegP and the different partners of the BAM complex in OM biogenesis in Salmonella. Our results show that the periplasmic chaperone SurA, unlike DegP, plays a major role in OMP biogenesis. This is in line with the results of Sklar et al. (2007b), suggesting that SurA is the primary chaperone responsible for the delivery of most OMPs to the OM in E. coli, while DegP plays a more discrete role in OMP assembly. The role of the BAM complex in OM biogenesis, as described for E. coli, is relatively well conserved in S. Enteritidis. In the culture conditions tested, BamC played no major role in OMP assembly, whereas BamE showed a slight involvement in the process, as described in E. coli, and this was consistent with an increased sensitivity of the S. Enteritidis ΔbamE mutant to Rf, which is a small toxic compound (this study and Sklar et al., 2007a). In contrast, BamD was shown to be very important for correct targeting of the OMP in S. Enteritidis. Interestingly, we found that BamD was not essential in Salmonella, unlike its homologue in E. coli (Malinverni et al., 2006; Onufryk et al., 2005), and its homologue in Neisseria gonorrhoeae, in which total deletion of comL (the bamD homologue) failed (Fussenegger et al., 1996). This constitutes the main difference between the BAM complex of Salmonella and that of E. coli.

Our results on OMP assembly and T3SS expression show clearly that a defect in OMP targeting is not responsible for a downregulation of T3SS expression in S. Enteritidis. Indeed, a ΔsurA mutant showed a marked reduction in the levels of the major OMPs in the OM, but showed no alteration in the secretion of T3SS-1 effectors and flagella proteins (Fig. 5). In a similar way, a ΔbamE mutant expressed T3SS to the same extent as the WT, while it presented a slight defect in OMP assembly. These results are in line with those of Lewis et al. (2008) who recently published that the loss of bamE results in a higher susceptibility to detergents and Rf indicative of altered OM integrity without showing any difference of T3SS-1-dependent secretion between their bamE mutant and its WT parent. It is interesting to note that surA and bamE mutants of Salmonella Typhimurium and S. Enteritidis have both been shown to have a slight decrease in their ability to enter eukaryotic cells (this paper and Humphreys et al., 2003; Lewis et al., 2008; Sydenham et al., 2000). As T3SS-1 and flagella seem to be produced at the correctly expressed levels in these strains, the invasion defect observed is probably related to the increased susceptibility of the strains to host-cell defences.

Results obtained with the ΔsurA mutant are all the more interesting because SurA has recently been suggested to deliver the OMP directly to the BAM complex (Sklar et al., 2007b), and to share a common function with BamB in the OMP assembly process (Ureta et al., 2007). This strongly suggests that the control of T3SS secretion observed in our ΔbamB mutant, and also probably in our bamD mutants, is not related to their OMP biogenesis defect, and raises the question of how the BAM complex, or some of the proteins of this complex, control T3SS secretion in Salmonella. One hypothesis is that control could be the result of the activation of the envelope stress response system σ^E observed in bamB and bamD mutants (Fardini et al., 2007; Onufryk et al., 2005). However, it is well known in E. coli that surA deletion results in a σ^E activation that is higher than that resulting from bamB inactivation. Therefore, σ^E is also likely to be activated in the S. Enteritidis ΔsurA mutant because of the observed OMP assembly defect. As T3SS secretion was not impaired in this mutant, it seems unlikely that the σ^E response, previously observed by microarray analysis in the S. Enteritidis ΔbamB mutant (Fardini et al., 2007), would be involved in the downregulation of the T3SS expression observed in the latter mutant or in the bamD mutants. The same reasoning could be followed for the bamE mutant, for which Lewis et al. demonstrated the activation of σ^E without observing a downregulation of T3SS secretion (Lewis et al., 2008). Additional arguments against the hypothesis that the activation of the σ^E pathway plays a role in the downregulation of T3SS secretion include the low-virulence Salmonella phenotype associated with the deletion of rpoE (σ^E-encoding gene) rather than with its overexpression (Humphreys et al., 1999), and the requirement of an intact σ^E for type-III secretion in Yersinia pseudotuberculosis (Carlsson et al., 2007).

Overall, our results demonstrate that the lack of OMP assembly does not systematically correlate with a downregulation of T3SS secretion in Salmonella. The results obtained using bamB and bamD mutants strongly suggest a more specific mechanism for the control of T3SS secretion. We are currently investigating a number of hypotheses. First, one can imagine a role of a complex composed only of BamD, BamB and probably BamA proteins, since no interaction has been demonstrated between BamD and BamB (Malinverni et al., 2006). Second, the BAM complex could process an unknown factor, independent of SurA, that could be responsible for the control of the T3SS secretion. One of the candidates could be the major OM protein component of T3SS: the secretins. Indeed, the missassembly of these components in the OM of bamB and bamD mutants could result in a feedback control leading to downregulation of the transcription of T3SS-related genes. However, this hypothesis seems to be unlikely, as we previously demonstrated that the bamB mutant was able to assemble functional T3SS-1 in its membrane when transcription of SPI-1 genes was forced by expressing HilA constitutively. Moreover, we did not observe a feedback control of SPI-1 gene transcription in a S. Enteritidis invA mutant unable to assemble T3SS-1 (Fardini et al., 2007). Finally, a role of BamB and BamD independent of their role in OM biogenesis could be envisaged. This is in line with the recent work of Khairnar et al. showing that BamB is involved in repair of DNA strand breaks, and homologous recombination, in E. coli. Those authors also demonstrated that BamB possesses a Ser/Thr kinase activity that is enhanced by pyrroloquinoline-quinone (Khairnar et al., 2007); therefore, BamB could be involved in one or more signalling cascades controlling at least this DNA process.

We are grateful to V. Legrand for her technical help. We thank Dr Roland Lloubes for providing the rabbit polyclonal anti-OmpA antisera. This work was supported in part by the European Union program FOOD-CT-2003-505523. Y. F. holds an INRA/Région Centre fellowship.

Edited by: W. Bitter