Contributed equally to this work. Present addresses: ‡Groupe à 5 ans Immunité Innée and Signalisation, Institut Pasteur, 75724 Paris Cedex 15, France; §Laboratoire de Microbiologie INSERM U411, Faculté de Médecine Necker-Enfants Malades, 75730 Paris Cedex 15, France; ¶Department of Molecular Biology, Umeå University, 901 87 Umeå, Sweden.
Signature-tagged mutagenesis (STM) was used to identify new genes involved in the virulence of the Gram-positive intracellular pathogen Listeria monocytogenes. One of the mutants isolated by this technique had the transposon inserted in virR, a gene encoding a putative response regulator of a two-component system. Deletion of virR severely decreased virulence in mice as well as invasion in cell-culture experiments. Using a transcriptomic approach, we identified 12 genes regulated by VirR, including the dlt-operon, previously reported to be important for L. monocytogenes virulence. However, a strain lacking dltA, was not as impaired in virulence as the ΔvirR strain, suggesting a role in virulence for other members of the vir regulon. Another VirR-regulated gene is homologous to mprF, which encodes a protein that modifies membrane phosphatidyl glycerol with l-lysine and that is involved in resistance to human defensins in Staphylococcus aureus. VirR thus appears to control virulence by a global regulation of surface components modifications. These modifications may affect interactions with host cells, including components of the innate immune system. Surprisingly, although controlling the same set of genes as VirR, the putative cognate histidine kinase of VirR, VirS, encoded by a gene located three genes downstream of virR, was shown not to be essential for virulence. By monitoring the activity of VirR with a GFP reporter construct, we showed that VirR can be activated independently of VirS, for example through a mechanism involving variations in the level of intracellular acetyl phosphate. In silico analysis of the VirR-regulated promoters revealed a VirR DNA-binding consensus site and specific interaction between purified VirR protein and this consensus sequence was demonstrated by gel mobility shift assays. This study identifies a second key virulence regulon in L. monocytogenes, after the prfA regulon.
Sensing and responding to environmental changes is a critical requirement for bacterial adaptation and survival in diverse conditions. Pathogenic bacteria in particular have to be able to respond to different stresses within the host, during the course of infection. This adaptive response is often under the control of two-component signal transduction systems (Msadek et al., 1993; Hoch, 2000). Typically, an external signal is sensed by the first component, a histidine kinase protein, which is autophosphorylated on a conserved histidine residue. The phosphate group is then transferred to a second protein that thereby becomes activated and in most cases binds to a DNA sequence. Such a set-up enables the cell to respond to various signals and regulate its gene expression accordingly.
The Gram-positive bacterium Listeria monocytogenes is the causative agent of listeriosis, a severe food borne disease affecting primarily newborns, pregnant women or immunocompromised individuals (Vazquez-Boland et al., 2001). L. monocytogenes is a facultative intracellular bacterium that is able to induce its own internalization into non-professional phagocytic cells. The molecular mechanisms underlying the infectious process of this bacterium have been well studied, allowing the identification of several virulence proteins (Cossart and Lecuit, 1998; Dussurget et al., 2004). Two invasion proteins, InlA and InlB, promote the entry of L. monocytogenes into various cell types. The bacterium escapes from the formed vacuole using a haemolysin, listeriolysin O (LLO), and a phopsholipase, PlcA. L. monocytogenes is able to spread from cell to cell, using actin polymerization driven by the protein ActA. Once in another cell, the secondary vacuole is lysed by LLO and another phospholipase, PlcB. All of these virulence factors are under the control of a key transcriptional regulator, PrfA, which is the only pleiotropic virulence regulator of L. monocytogenes hitherto found. PrfA expression itself is thermo-regulated by an RNA thermosensor which allows maximal expression of virulence genes at 37°C (Johansson et al., 2002). The complete genome sequence of L. monocytogenes (Glaser et al., 2001) suggests that this bacterium harbours 16 two-component systems among which five have been studied further: CheY/CheA, involved in chemotaxis (Dons et al., 1994), LisR/LisK, implicated in acid stress resistance (Cotter et al., 1999; 2002), AgrA/AgrC (Autret et al., 2003), CesR/CesK (Kallipolitis et al., 2003) and the orphan response regulator DegU (Knudsen et al., 2004) indirectly involved in virulence. Moreover, using insertional inactivation, Kallipolitis et al. had shown that seven listerial two-component systems contribute to pathogenesis and stress tolerance of the bacterium (Kallipolitis et al., 2003).
We applied signature-tagged mutagenesis (STM) in order to identify new genes involved in the virulence of L. monocytogenes. One of the mutants identified by this technique had the transposon inserted in a gene encoding a response regulator of a two-component system named VirR, for Virulence Regulator. In the present work we present evidence that this response regulator is important for virulence of L. monocytogenes and that the genes it regulates are mostly involved in bacterial surface components modifications.
Identification of the virR gene by STM and analysis of the vir locus
We applied STM on L. monocytogenes strain EGD (serotype 1/2a), using a Tn917 derivative transposon, to create a bank of mutants. Mutants were screened in vivo for reduced colonization of the spleen and the liver of female Swiss mice, 72 h after i.v. infection (Dramsi et al., 2004). One of the mutants isolated was used for the study described in this article. Southern blot analysis showed that the transposon insertion in the chromosome was unique (data not shown). The DNA sequences flanking the transposon were analysed, revealing that the Tn917 had inserted in the middle of the lmo1745 gene (Glaser et al., 2001). lmo1745 encodes a protein of 225 amino acids exhibiting a high degree of homology to response regulators of the OmpR-PhoB family. Due to the reduced virulence of the mutant in mice, it was named virR for Virulence Regulator. The virR gene is the third gene in a putative polycistronic operon (Fig. 1) consisting of seven genes. The two genes upstream of virR, lmo1747 and lmo1746, show homologies to an ATP-binding protein and a permease of an ABC transporter system whose substrate is unknown. lmo1744 and lmo1743, located immediately downstream of virR, show no similarity to genes encoding proteins of known functions. The third gene downstream of virR, adeC, is very similar (67% homology) to the adenine deaminase gene of the same name in Bacillus subtilis (Nygaard et al., 1996). In this bacterium, AdeC is implicated in the use of adenine as a carbon and nitrogen source. The last gene of this putative operon encodes a protein that shows strong homologies to bacterial histidine kinases of two-component systems. Our working hypothesis was that this gene could encode the cognate histidine kinase of VirR, and it was therefore designated virS. The virS gene is 1041 bp long and encodes a protein of 346 aa (≈ 41 kDa). VirS is predicted to contain two transmembrane segments in its N-terminal part (positions 12–30 and 40–62 of the amino acid sequence).
VirR, but not VirS, is required for virulence of L. monocytogenes
To investigate the role of VirR and VirS during virulence, a non-polar virR deletion mutant (ΔvirR) was generated by an in frame deletion of the virR gene. virS was inactivated by deleting 200 bp in the middle of the gene (see Experimental procedures). Growth of the virR::Tn917, ΔvirR and the ΔvirS mutants in brain heart infusion (BHI) showed no significant difference when compared with the wild-type EGD strain (data not shown). The virulence of the mutant strains were compared with the wild-type EGD strain by determining the LD50 of these strains after intravenously infecting BALB/c mice. The LD50 of the virR::Tn917 strain [6.7 × 105 colony-forming units (cfu)] was much higher than the LD50 of the EGD (1.1 × 104 cfu), confirming the result from the STM-analysis that this mutant was affected in virulence. The LD50 of the ΔvirR strain (2.7 × 106 cfu) was similar to the LD50 of the virR::Tn917 mutant and much higher than the LD50 of the EGD strain. Surprisingly, the ΔvirS mutant (1.1 × 104 cfu) showed no difference in LD50 as compared with the wild-type EGD strain. To analyse if the reduced virulence of the virR transponson mutant strain was due to a polar effect, a deletion mutant of the three genes downstream virR, i.e. lmo1744, lmo1743 and adeC (strain Δ1744-adeC,Fig. 1) was generated. This strain showed no significant difference in LD50 (5.3 × 104 cfu) compared with the EGD wild-type strain, clearly demonstrating that the defect in virulence for the virR::Tn917 strain was due to the inactivation of the virR gene by the transposon and not to a polar effect on the other downstream genes. This was also confirmed by other results (see below). In conclusion, inactivation of the virR gene, but not that of the virS gene, affects virulence.
Reduced in vivo growth of a strain lacking VirR
To further compare the virulence of the wild type (EGD) and ΔvirR strain and document the behaviour of the bacteria, we assessed the ability of these bacteria to colonize and survive in the spleen and liver of BALB/c mice over a 3 day period after intravenous injection (5 × 103 bacteria per mouse). The results obtained showed that the ΔvirR strain has a reduced ability to grow and multiply within both liver and spleen (Fig. 2A and B). After 72 h, the number of cfu of the wild type (EGD) was between 100- and 1000-fold higher when compared with the cfu for the ΔvirR strain detected in the spleen and liver respectively. These results further support a crucial role for VirR in the establishment of a successful L. monocytogenes infection.
Both the virR and virS mutants are affected in entry into Caco2 cells
To further analyse the role of the VirR/VirS two-component system at the cellular level, we studied the capacity of the virR and virS mutants to invade Caco2 cells. The virR::Tn917 strain, as well as the ΔvirR strain, showed a remarkable decrease in entry into Caco2 cells compared with the EGD strain (about 3% of the rate of entry of the wild-type strain for both mutants), as assessed by a gentamicin survival assay (Fig. 3A). The ΔvirS mutant also showed a reduced entry (8% of the rate of entry of EGD). In comparison, the Δ1744-adeC strain (lacking the genes in between virR and virS) presented no difference in entry compared with EGD, again showing that a putative polar effect caused by the transposon insertion did not affect virulence.
We also investigated whether the mutants were affected in adherence by determining the number of bacteria associated with Caco2 cells 1 h after infection. The number of cell-associated bacteria decreased for both the virR::Tn917 strain (twofold) and the ΔvirS mutant strain (fivefold) when comparing with the wild-type EGD strain (Fig. 3B). These results suggest that the VirR/VirS system is important for adhesion and entry of L. monocytogenes to eukaryotic cells. As entry of L. monocytogenes into Caco2 cells has been shown to be dependent on the InlA protein (Gaillard et al., 1991), we analysed the presence of InlA at the surface of the ΔvirR and ΔvirS mutants by immunofluorescence. InlA was present at the surface of both mutants at the same level and with the same distribution as in the wild-type strain (data not shown), suggesting that other factors regulated by VirR/VirS are implicated in the entry process.
Entry of the virR mutant in Caco2 cells can be partially complemented
We wanted to confirm that VirR alone is responsible for the observed effect on virulence. Therefore, we complemented the mutant by a plasmid, pvirR, derived from pP1 (Dramsi et al., 1995) harbouring the entire virR gene under the control of the constitutive and strong promoter Pprot (see Experimental procedures). Invasion of Caco2 cells could be partially restored with the resulting ΔvirR–pvirR complemented strain as compared with the control ΔvirR-pP1 strain (Fig. 4). However, we were unable to follow the growth of the ΔvirR–pvirR and the control strains in the organs of BALB/c mice after intravenous inoculation as bacteria rapidly lost the plasmid in the absence of selection pressure after inoculation in the animals (data not shown).
Identification of the genes regulated by VirR
Based on its amino acid sequence, VirR belongs to a family of response regulators possessing a helix-turn-helix domain, and is most probably a transcriptional regulator activating and/or repressing a set of genes upon activation by its cognate histidine kinase. We therefore investigated which genes are regulated by the VirR/VirS system using a transcriptome approach. For this purpose, whole-genome arrays containing 95% of the 2853 predicted open reading frames of L. monocytogenes were used (Milohanic et al., 2003). The expression profiles of the virR::Tn917 and the ΔvirS mutants were compared with those of strain EGD. Twofold expression differences or more were taken into account (Experimental procedures).
Figure 5 shows the level of transcription of each gene in the virR::Tn917 strain compared with the wild-type EGD strain after growth in BHI. Only 17 genes showed a significantly different level of expression when comparing both strains. All of these genes were more transcribed in the wild-type strain than in the virR::Tn917 mutant (Table 1) and hence most likely activated by VirR. Of these 17 genes, five corresponded to virR itself and the four genes immediately downstream virR. The lower expression levels of these genes in the virR::Tn917 mutant are probably due to a polar effect of the transposon insertion, although this polar effect does not contribute to the reduced virulence of this strain as demonstrated above.
Table 1. Genes differentially regulated in the wild-type L. monocytogenes EGD relative to EGD virR::Tn917.
Unknown, similar to ABC transporter (ATP-binding protein)
d-alanine-activating enzyme (dae), d-alanine- d-alanyl carrier protein ligase (dcl)
DltB protein for d-alanine esterification of lipoteichoic acid and wall teichoic acid
d-alanyl carrier protein
DltD protein for d-alanine esterification of lipoteichoic acid and wall teichoic acid
Similar to MprF protein of S aureus
Similar to unknown protein
Similar to unknown protein
virR and downstream genes
Two-component system response regulator
Unknown, similar to unknown proteins
unknown, highly similar to adenine deaminases
Two-component system histidine kinase
The 12 other genes are organized in seven predicted transcriptional units (three operons and four single genes). Four genes encode proteins of unknown functions or similar to proteins from other bacteria. Two genes, forming a putative operon (lmo2114, lmo2115), encode the permease and the ATP-binding protein of an ABC transporter respectively. Two other genes, lmo1696 and lmo1695, encode a protein containing a VanZ domain and a protein very similar (> 38% identity) to the MprF factor identified in Staphylococcus aureus (Peschel and Collins, 2001) respectively. VanZ proteins confer low-level resistance to the glycopeptide antibiotic teicoplanin (Arthur et al., 1995). MprF, also known as FmtC, is a lysinyl-transferase which covalently binds a lysine residue to phosphatidyl-glycerol in the membrane and has been shown to confer resistance to human defensins and cationic peptides in S. aureus (Staubitz et al., 2004). The four remaining genes belong to the dlt operon (dltA, dltB, dltC, dltD). The dlt operon encodes the proteins necessary for d-alanylation of cell wall teichoic and lipoteichoic acids (LTAs) (Neuhaus and Baddiley, 2003) (Perego et al., 1995). Interestingly, the dlt operon has been shown to play a critical role in adhesion and virulence of the L. monocytogenes strain LO28 (Abachin et al., 2002).
Identification of the genes regulated by the VirS histidine kinase
In order to investigate that VirS was the cognate kinase of VirR, a transcriptomic analysis of the ΔvirS strain was performed and compared with the expression levels of the parental EGD strain. As expected, transcriptomic analysis of the ΔvirS strain showed that the 12 genes identified as being VirR-regulated were also less expressed in the virS mutant strain, confirming that VirS and VirR are members of the same two-component system. More surprisingly, 108 additional genes were shown to be differentially expressed in the ΔvirS strain as compared with the virR::Tn917 strain, with 91 of them being more expressed in the wild-type strain and 29 less expressed (Supplementary material, Table S1). Nearly one-third (29%) of the genes identified encoded proteins of unknown function. The other VirS controlled genes encoded proteins known or predicted to be involved in, metabolism (30%), regulation (13%), transport (14%), stress response (8%), virulence (1%) or in other functions (5%) (see Supplementary material, Table S1). Thus, VirS controls a broader set of genes, most probably by cross-talking with one or more response regulators.
Analysis of a dltA mutant
As the dlt-operon had previously been reported to be required for virulence in the serotype 1/2c strain LO28 (Abachin et al., 2002), we investigated the role of the dlt-operon in virulence in the EGD strain background (serotype 1/2a). An EGD strain deleted for the dltA gene was constructed (see Experimental procedures) in order to assess its virulence abilities in vivo and determine to which extent regulation of dlt by VirR is critical for virulence. The ΔdltA mutant strain showed a reduced ability to colonize and survive within the liver and spleen as compared with the wild type (100- and fivefold less bacteria after 72 h respectively) (Fig. 2). When examining the liver and spleen at earlier time points (24 and 48 h, respectively), no difference was seen between the EGD wild-type strain and the ΔdltA knock-out strain. This is in strong contrast to the ΔvirR knock-out strain, which already after 24 h showed a reduced ability to colonize these organs (Fig. 2). Interestingly, as shown in Fig. 4, the ΔdltA-mutant strain was still able to invade Caco2 cells equally well as the EGD wild-type strain, and consequently, to a much higher degree than the ΔvirR strain. Taken together, these findings show that although the dlt-operon is important for virulence, the regulation of the genes within the dlt-operon is not the only factor contributing to the virulence of the virR mutant strains.
Identification of a palindromic sequence upstream of the VirR-regulated genes
An in silico approach was taken to search for DNA consensus motifs present within the promoters of all VirR-regulated genes (i.e. lmo2114–2115, dltABCD, lmo2439, lmo2156, lmo2177, lmo0604, lmo1696–1695). A palindromic consensus motif of 16 bases, CTNACAwwwTGTNAG, was identified for all the regulated promoters (Fig. 6A). For five of the promoters of these transcriptional units (lmo2439, lmo2156, dltABCD, lmo2114–2115 and lmo2177), the palindromic sequence was located exactly 21 bp upstream of the putative −10/−35 box (Fig. 6A). The transcriptional data, together with the in silico data, therefore suggest that this consensus motif constitute the VirR DNA-binding site.
VirR binds specifically to the palindromic sequence found in the promoter region of VirR-regulated genes
In order to demonstrate binding of VirR to the palindromic sequence, a gel mobility shift assay was performed. A VirR recombinant protein was purified (see Experimental procedures) and subsequently incubated with a 50 bp labelled DNA fragment corresponding to the palindromic sequence found within the lmo2156 promoter region (Table 2). As shown in Fig. 6B, binding of VirR to this sequence formed a DNA–protein complex migrating slower than the uncomplexed DNA on a native poly acrylamide gel (lane 1–4). With low VirR concentrations, a single complex was formed (marked with an asterisk), whereas two complexes were detected when a larger amount of VirR was used (marked with two asterisks). A 10 times excess of the same, non-radioactive, DNA-fragment was able to out-compete and completely abolish the binding of VirR (lane 5), whereas a 10-fold excess of another non-radioactive unspecific DNA fragment (SBD, Table 2) was essentially unable to alter the binding of VirR to the DNA containing the palindromic sequence (lane 6). These results clearly demonstrate that VirR binds specifically to the palindromic sequence found in the promoter region of lmo2156, and probably to the seven other VirR-regulated operons.
Table 2. Bacterial strains, plasmids and oligonucleotides used in this study.
Underlined nucleotides in the oligonucleotide sequences indicate restriction sites.
VirR regulation of the expression of the dlt operon
In order to follow the activation of VirR in Listeria, we constructed a plasmid, derived from pNF8 (Fortineau et al., 2000), containing a gfp gene under the control of the dlt promoter (pPdlt-gfp,Table 2). Cells were first examined under the microscope in BHI, at mid-exponential phase (OD600 = 0.6). As expected, EGD cells carrying this plasmid were all fluorescent, although heterogeneity in the level of GFP fluorescence could be observed (Fig. 7A). The virR::Tn917 strain harbouring the pPdlt-gfp plasmid was not fluorescent, demonstrating that VirR is necessary for activation of the transcription of the dlt operon. Surprisingly, the ΔvirS strain containing the pPdlt-gfp plasmid appeared very heterogeneous, a major part of the bacterial population being not or weakly fluorescent whereas a small fraction of the bacteria was fluorescing at levels comparable with the wild-type strain. Segregational instability of the plasmid could not explain this phenotype because cells were cultured in the presence of a selective antibiotic (chloramphenicol). Furthermore, when the ΔvirS strain harbouring the pPdlt-gfp plasmid was grown in BHI containing chloramphenicol until mid-log phase, and thereafter plated on agar with and without chloramphenicol, all the bacteria still expressed the resistance marker, confirming that loss of the plasmid had not occurred. Thus, the most likely explanation of the GFP expression in a fraction of the ΔvirS population is that VirR can be activated independently of its cognate kinase, VirS.
VirR can be activated independently of VirS in presence of acetate
It has been demonstrated that acetyl phosphate can phosphorylate and activate several response regulators (such as CheY, PhoB or OmpR) in vitro (McCleary and Stock, 1994), although the role of this activation in vivo remains unclear. We investigated whether such a mechanism could be responsible for the partial activation of VirR in the ΔvirS strain. As shown above, because VirR is able to activate transcription of dlt, we used the pPdlt-gfp plasmid directly as a reporter of the activation of VirR inside the bacteria. When accumulating inside the bacteria, acetate is transformed into acetyl phosphate by the acetate kinase (McCleary and Stock, 1994). We used this property to increase the amount of acetyl phosphate in the bacteria, and inoculated bacteria in minimal medium (Fortineau et al., 2000) containing acetate or glucose (control) as carbon sources. As expected, EGD harbouring the pPdlt-gfp plasmid was brightly fluorescent in minimal medium containing either carbon sources as observed when this strain was grown in BHI (data not shown). Interestingly, the number of fluorescent ΔvirS+ pPdlt-gfp cells, showed a fourfold increase (from 3% to 12% of the bacterial population) in presence of acetate as compared with glucose. Thus, VirR can be activated in presence of acetate in the medium, most likely via production of acetyl phosphate inside the bacteria.
In this work, we describe the identification and characterization of a new two-component system, VirR/VirS, involved in L. monocytogenes virulence. As revealed by a transcriptomic approach, the response regulator VirR positively controls the transcription of 12 genes. Among those, we found the dlt-operon which is involved in d-alanylation of cell wall teichoic acids (TAs) and LTAs. It has already been shown that the dlt-operon plays an important role for virulence in the L. monocytogenes strain LO28 (serotype 1/2c) (Abachin et al., 2002). In particular, it was shown that a dltA mutant in this background was severely affected in its adhesion to eukaryotic cells, as observed in this report for both the virR and virS mutants. However, we show here that the ΔdltA strain in an EGD background (serotype 1/2a), was less affected in its ability to spread and multiply within the spleen and liver of BALB/c mice than the ΔvirR isogenic strain (Fig. 2). Moreover, the EGDΔdltA strain was as efficient in entry into the Caco2 eukaryotic cell line as the parental EGD strain, as demonstrated by cell-culture invasion experiments (Fig. 4). Thus, our results clearly demonstrate that reduced expression of the dlt-operon in the virR mutant only partially explains the reduced virulence of this strain. The difference of virulence among dlt mutants in different L. monocytogenes serotypes highlights the absolute necessity to compare isogenic strains. VirR also activates the expression of a gene showing significant homology (39%) to the MprF protein of S. aureus and other bacterial species. Esterification of cell surface components with amino acids, such as lysinylation of phosphatidyl-glycerol (a major lipid of the bacterial membrane) by MprF, and also d-alanylation of cell wall TAs and LTAs by the dlt complex lead to a reduced negative charge of the bacterial surface. One of the consequences of this process is the repulsion of cationic peptides such as human defensins, which explains the susceptibility of dlt or mprF mutants of S. aureus to these bactericidal compounds (Peschel et al., 1999) (Peschel and Collins, 2001). Regulation of both dlt and mprF by VirR strongly suggests that one main role of the VirR/VirS system is to regulate resistance of L. monocytogenes to human defensins or to cationic peptides. Furthermore, a mass of recent reports has highlighted the role of surface components, e.g. LTAs and peptidoglycan, in the innate immune response through Toll-like receptors (Takeda and Akira, 2005). We speculate that the lack of LTAs modification in the virR mutant increases the innate immune response through TLR2 resulting in the earlier elimination of bacteria.
Inactivation of the virR gene severely affected virulence as the LD50 of ΔvirR and that of virR::Tn917 was almost 100-fold increased compared with the wild-type strain. Surprisingly, deletion of the virS gene, encoding the cognate histidine kinase of VirR, had no similar effect on virulence, because the ΔvirS strain showed no difference in LD50 as compared with the wild-type strain. When using a GFP reporter system, we could observe that a fraction of a ΔvirS population remained fluorescent during growth, indicating that VirR could be activated in these cells despite the absence of its cognate kinase. Thus, we propose that inside the host, VirR may be activated by a mechanism independent of VirS, which would explain the lack of attenuation of the ΔvirS strain. Activation of VirR may occur either by another histidine kinase cross-talking with VirR, or via a molecular compound phosphorylating the response regulator. In agreement with this hypothesis, we have been able to modulate the activation of VirR in the ΔvirS cells using acetate as a precursor of acetyl phosphate. It has been proposed that acetyl phosphate modulates the activation of response regulators, rather than functioning as an activator per se (McCleary and Stock, 1994). In the ΔvirS strain, with the cognate kinase missing, acetyl phosphate may act as a donor of the phosphate group to VirR. It may be the case when bacteria infect the host and/or reside within the host cell, allowing activation of VirR independently of VirS. Interestingly, the ΔvirS strain, like the ΔvirR strain, was affected in entry into Caco2 cells. However, entry into cells may vary according to the cell lines used and invasion assays are only indicative of possible effects on virulence. In contrast, animal studies in adequate models are the best assays to address the role of a gene in virulence. Thus, we hypothesize that the apparent discrepancies in between invasion assays and animal studies come from the activation of VirR independently of VirS. This activation mechanism would also explain why the expression of the dlt-operon is stochastic (i.e. only affecting an apparently undetermined fraction of the bacterial population) in the ΔvirS strain, as seen using our GFP reporter construct (Fig. 7). Stochastic phenomena in bacterial populations can often be explained by variation of the concentration of an activating compound in individuals of a population (Swain et al., 2002). Thus, the stochastic expression of GFP in the bacterial population could be explained by the variation in concentration of acetyl phosphate among the cells. Finally, although unlikely, we cannot exclude that among the vast amount of genes affected in the ΔvirS strain, one or more gene could act as virulence inhibitors, as it has been shown for patho-adaptive genes (Sokurenko et al., 1999). The lack of expression of such a gene would compensate for the reduced virulence in the ΔvirS strain.
In almost all cases, genes encoding two-component systems are contiguous on the chromosome. This is not the case for virR and virS because three genes separate them. Such an organization has not been observed in other bacteria, except for the DevS/DevR system of Mycobacterium smegmatis (O’Toole et al. 2003). Importantly, all of the genes found regulated by VirR were also found controlled by VirS, clearly demonstrating that VirS and VirR are members of the same two-component system. However, the transcriptomic analysis of both the virR::Tn917 and ΔvirS mutant strains revealed that VirR positively controlled 12 genes (seven transcriptional units, Table 1), whereas VirS regulated 120 genes in both a positive and a negative manner (Supplementary material Table S1). The simplest explanation of this difference would be that the VirS histidine kinase is able to crosstalk with one or several other response regulators, leading to the activation/repression of the transcription of these genes. Further work is required to shed light on this apparent peculiarity concerning VirR/VirS and their partner(s).
By an in silico approach, we have identified a putative VirR DNA-binding site. This palindromic region of 16 bases was found upstream of all the seven transcriptional units commonly regulated by VirR and VirS. The palindrome was highly conserved between the different promoter regions and showed a common organization, strengthening its importance in determining the specificity of the binding of VirR to the DNA. Furthermore, we also found duplicates of the palindrome in some of the promoters, although the sequences of these alternative sites showed weaker similarity to the established consensus (data not shown) and their putative function still remains elusive. In strong agreement with our predictions, we indeed observed specific binding of VirR to the palindromic sequence contained in the lmo2156 promoter region when performing gel shift experiments with purified VirR protein (Fig. 6B). We also found that the level of activation, as given by the transcriptomic approach, was higher in promoters displaying a VirR-binding site more similar to the consensus (Table 1 and Fig. 6A). Thus, our results indicate that the VirR protein activates transcription directly by binding the palindromic sequence upstream of the target genes.
Interestingly, by searching the Listeria genome (Glaser et al., 2001), we found only one other sequence homologous to the vir palindrome. This 16 bp palindrome was located within a 246 bp intergenic region between two divergent genes: lmo0858, encoding a protein similar to a transcription regulator of the LacI family and lmo0859, encoding a periplasmic sugar-binding protein of a putative ABC transporter. These two genes were not found as VirR-regulated using the macroarrays, indicating either the limit of the technique, or that another yet unknown gene, possibly a-non-coding RNA, regulated by VirR is located inside the intergenic region itself. This issue is under investigation.
In a very recent study, Williams et al. analysed deletion mutants of 15 out of the 16 response regulators of L. monocytogenes and detected an effect on virulence only for the ΔdegU mutant. These results are not in agreement with the results obtained for CheY/CheA (Dons et al., 1994), LisR/LisK (Cotter et al., 1999; 2002), AgrA/AgrC (Autret et al., 2003), CesR/CesK (Kallipolitis et al., 2003) and the VirR/VirS system described in this article. These discrepancies will require further investigation.
In conclusion, up to now, the PrfA regulon was the only known virulence regulon of L. monocytogenes. This work identifies a second virulence regulon involved in listerial pathogenesis. Further work is required to identify the other gene(s) controlled by VirR, which is/are necessary for L. monocytogenes virulence.
Bacterial strains and growth conditions
The strains and plasmids used in this study are described in Table 2. Escherichia coli was grown in Luria–Bertani medium (LB, Difco Laboratories), L. monocytogenes was grown in BHI (Difco Laboratories) at 37°C and was maintained on BHI agar. Stock cultures were stored at −80°C. BHI medium was buffered to a pH of 5.5 with MES [100 mM 2-(N-morpholino) ethanesulphonic acid]. Minimal medium was synthesized as previously described (Phan-Thanh and Gormon, 1997). When appropriate, antibiotics were added at the following concentrations: 100 µg ml−1 of ampicillin for E. coli and 5 µg ml−1 of erythromycin, 7 µg ml−1 of chloramphenicol for L. monocytogenes.
General DNA methods
Plasmid DNA was isolated using the Qiagen Plasmid Kit (Qiagen) and polymerase chain reaction (PCR) products were purified using the PCR Purification Kit (Qiagen). Restriction enzymes were purchased from Boehringer. Ligations were performed overnight at 4°C using the T4 DNA Ligase from Boehringer. PCR were performed on a thermocycler (MJ Research). Transformation of L. monocytogenes was performed according to a previously described protocol (Park and Stewart, 1990). Plasmid pPM30 was constructed by, into pMK4 (Sullivan et al., 1984), inserting a 431 bp EcoRI-BamHI fragment containing the Pdlt promoter of L. monocytogenes from the pNF8 plasmid (Fortineau et al., 2000).
For each deletion, genomic DNA was extracted from L. monocytogenes EGD and used as a template to amplify two different fragments, one derived from the 5′ part and one from the 3′ part of the genes to be deleted with oligos containing the appropriate restriction sites (see Table 2). The SOE (splicing by overlap extension) PCR procedure (Horton et al., 1990) was used to splice the PCR products from either side of the sequence to be deleted. The hybrids were subsequently cloned into the appropriate sites of the temperature-sensitive shuttle vector pKSV7 for virS, yielding plasmid pHF33, or in the pMAD vector (Arnaud et al., 2004) for the other deletions. Plasmids were electroporated into the L. monocytogenes EGD strain at 2500 V, 200 Ω and 25 µf. Transformants were selected at 30°C on BHI agar plates containing chloramphenicol. Allelic exchange was achieved as previously described (Arnaud et al., 2004) by selecting loss of the plasmid after successive steps at permissive/non-permissive temperature. Gene replacement was verified by PCR and sequencing of the PCR product using the oligos located at the 5′ end of the upstream fragment previously amplified with the oligo located at the 3′ end of the downstream fragment (see Table 2).
Complementation of the ΔvirR strain
A DNA fragment containing the virR gene was amplified by PCR using oligos VircA and VircB (Table 2). The PCR product was purified and subsequently digested using the appropriate restriction enzymes and cloned into the pP1 plasmid (Dramsi et al., 1995), giving rise to pvirR. The resulting plasmid was electroporated into the EGD ΔvirR strain at 2500 V, 200 Ω and 25 µf and bacteria were selected by plating on BHI agar containing erythromycin. Sequence of the resulting plasmid, pvirR was verified by sequencing the insert from both junctions.
Fifty per cent lethal dose (LD50) experiments were performed by injecting 4 to 6 week old specific pathogen free-female BALB/c mice (Charles River) intravenously with 0.3 ml of serial dilutions of bacteria. The LD50 was determined using the probit method.
Bacterial growth in mice was determined by injecting 4 to 6 week old specific pathogen free-female BALB/c mice (Charles River) with a sublethal dose of 5 × 103 cfu per bacterial inoculum. At 24 h, 48 h and 72 h post infection, mice were killed and spleen and liver were sterilely dissected. Number of cfu per organ was determined by plating tenfold serial dilutions of organs homogenates on BHI agar plates. For plasmid containing bacteria, organ homogenates were plated on BHI agar plates with or without antibiotic in order to determine stability of the plasmid.
Invasivity tests were performed using the gentamicin survival assay as previously described (Gaillard et al., 1991) (Mengaud et al., 1996). Briefly, human carcinoma cell line Caco-2 was propagated as previously described (Gaillard and Finlay, 1996). Caco-2 cells were added to wells at a density of 105 cells ml−1 in 24-well tissue culture plates and were incubated for 24 h at 37°C in 10% CO2. Listeria strains were grown to OD600 = 0.8, washed in PBS and diluted in DMEM such that the multiplicity of infection (moi) was about 15 bacteria per cell. Bacterial suspensions were added to the Caco-2 cells for 1 h. Cells were then washed twice and extra cellular non-invasive bacteria were killed by adding 10 µg ml−1 gentamicin for 1 h. After washing, cells were lysed in 0.2% Triton X-100, and the number of viable bacteria was assessed by plating serial dilutions on agar plates.
Images were acquired on a Zeiss Axiovert fluorescence 135 microscope (Zeiss, Germany), equipped with a cool charged device camera (micromax 512BFT, Princeton Instruments, NJ) driven by a Metamorph Imaging System software (Universal Imaging, Downingtown, PA). Appropriate filter set with narrow band pass to avoid fluorescence crossover were used to acquire fluorescent images. Oil immersion objectives (60× and 100×) were used.
Adhesion assay were performed as previously described (Milohanic et al., 2003). Briefly, bacteria were grown in BHI to OD600 = 0.8, washed twice in PBS and diluted appropriately in DMEM. Caco-2 cells cultured on cover slips were inoculated at an moi of ≈ 100 bacteria/cell. After 1 h incubation they were washed three times in PBS, fixed with Para formaldehyde 3% for 30 min and washed three times again in PBS before being processed for fluorescence labelling. Adherent bacteria were counted by examining > 250 cells in randomly picked microscopic fields.
Overnight cultures of bacteria were diluted appropriately in 10 ml of BHI-MES pH 5.5 and grown at 37°C with shaking to OD600 = 0.6. Bacteria were then harvested for 5 min at 6000 g at 4°C. For RNA extraction, cells were resuspended by vortexing in 400 µl of resuspension buffer (12.5 mM Tris, 5 mM EDTA, and 10% glucose). Five hundred microlitres of acid phenol (pH 4.6) and 0.4 g of glass beads (0.2–0.3 mm diameter, Sigma) were added and then the cells were sheared mechanically using a Fastprep apparatus (Bio 101). After centrifugation at 13 000 g for 5 min, the supernatant was transferred to a fresh tube and 1 ml of Trizol reagent (Gibco BRL) was added. The sample was incubated for 15 min at room temperature. Total RNA was extracted twice with chloroform iso-amyl-alcohol (24:1, v/v) and precipitated in 0.7 vol of isopropanol. After washing step with 70% ethanol, the RNA pellet was dissolved in sterile DEPC treated water and quantified by absorbance at 260 and 280 nm.
Macroarrays studies were performed as described previously (Milohanic et al., 2003). Briefly, total RNA samples (1 µg) were used as templates for production of 33P-labelled cDNA using a set of 3′ specific oligonucleotides and the AMV reverse transcriptase (50 U; Roche). Labelled cDNA was then purified on a QIAquick column (Qiagen). Whole genome macroarrays of L. monocytogenes EGD-e were prehybridized for 2 h in hybridization solution at 65°C, then hybridized with labelled cDNA for about 20 h at 65°C. After hybridization, membranes were washed twice at room temperature and twice at 65°C. Arrays were then sealed in thin bags and exposed to a phosphor screen (Molecular Dynamics) for 24–72 h. Membranes were scanned using a 445SI Phosphorimager (Molecular Dynamics) and analysed using the Arrayvision software (Imaging Research) as described (Milohanic et al., 2003). Statistical significant genes in a set of arrays were identified using the SAM technique (Tusher et al., 2001) (http://www-stat.stanford.edu/~tibs/SAM/). We chose a delta value corresponding to a false discovery rate < 10%. Only genes with a twofold expression change or more were taken into account. For each strain, at least two independent RNA preparations were tested, and two cDNAs from each of the RNA preparations were hybridized to two sets of arrays and analysed.
A 678 bp PCR fragment corresponding to the virR gene was produced using primers VirR-UP and VirR-down (see Table 2) with the Pwo polymerase (Roche). Subsequent PCR product was inserted into the pET101/D-TOPO vector (Invitrogen) as described by the manufacturer. Sequence of the resulting plasmid, pET-VirR was verified by sequencing the insert from both junctions. It was used to transform E. coli Tuner + (DE3) (Novagen). The resulting strain was grown in 500 ml of LB medium at 37°C and expression was induced at mid-exponential phase by addition of 1 mM of IPTG (isopropyl-β- d-thiogalactoside) and incubation pursued for 4 h. Cells were centrifuged at 5000 g for 20 min and the bacterial pellet was resuspended in 10 ml of B-Per reagent (Pierce). Purification was performed on a Ni-NTA agarose column (Qiagen) according to the manufacturer's instructions. Protein concentration was determined using the BCA Protein Assay (Pierce) according to manufacturer's instructions.
Gel mobility shift DNA-binding assay
Oligonucleotides 2156 and 2156-cplt (see Table 2) corresponding to the palindromic sequence contained in the promoter region of the lmo2156 gene were annealed at 95°C for 5 min. dsDNA was then labelled using the poly T4 kinase (Invitrogen) in presence of [γ-32P]- dATP (32 µCi). Binding of VirR to DNA was carried out in a reaction mixture containing 104 c.p.m. 32P-labelled DNA, 1 µg of poly (dI-dC) (Pharmacia), and 5:1 binding buffer (25 mM HEPES, pH 7.5, 0.1 mM EDTA, 5 mM DTT, 10% glycerol). The DNA-binding reaction was initiated by the addition of VirR and incubated at room temperature for 20 min. Samples were then loaded directly onto a 6% polyacrylamide gel (50 mM Tris, 400 mM glycine, 1.73 mM EDTA) during electrophoresis (14 V cm−1). Electrophoresis was pursued for 1 h at room temperature, and the gels were then dried and analysed by autoradiography.
Non-labelled oligonucleotides 2156/2156-cplt and SBD/SBD-cplt (see Table 2) were annealed at 95°C for 5 min for specific and non-specific competition respectively. These dsDNA fragments were added to the binding mixtures, prior to the addition of VirR.
We thank all members of Pascale Cossart's laboratory, especially Francis Repoila for helpful and entertaining discussions. This work was supported by the Pasteur Institute (GPH9), INSERM and the French Ministry of Research (Programme de Microbiologie Fondamentale et Appliquée; MIC 0312). Pascale Cossart is an international scholar from the Howard Hughes Medical Institute.