Environmental control of invasin expression in Yersinia pseudotuberculosis is mediated by regulation of RovA, a transcriptional activator of the SlyA/Hor family



Invasin is the primary invasive factor of Yersinia pseudotuberculosis that allows efficient internalization into eukaryotic cells. We investigated invasin expression and found that the inv gene is regulated in response to a variety of environmental signals, such as temperature, growth phase, nutrients, osmolarity and pH, and requires the product of rovA, a member of the SlyA/Hor transcriptional activator family. The rovA gene was found by a genetic complementation strategy that restores temperature regulation of an unexpressed inv–phoA fusion in Escherichia coli K-12. RovA plays a role in the invasion of Y. pseudotuberculosis into mammalian cells and mediates the regulation of invasin in response to all environmental signals analysed. Deletion analysis of the inv promoter region revealed a DNA segment extending 207 bp upstream of the transcriptional start site, which is required for maximal RovA-induced inv transcription. Gel retardation assays showed that RovA interacts preferentially with this promoter fragment and suggested two potential RovA binding sites. Studies with chromosomal gene fusions also demonstrated that rovA follows the same pattern of regulation as invasin, indicating that environmental control of inv expression is mainly mediated by the control of RovA synthesis. Furthermore, we showed that a rovA–lacZ fusion is only slightly expressed in a rovA mutant strain, indicating that a positive autoregulatory mechanism is also involved in rovA expression.


The two enteropathogenic Yersinia species, Yersinia enterocolitica and Yersinia pseudotuberculosis, cause different food-borne diseases with a broad spectrum of intestinal and extraintestinal manifestations, ranging from a self-limiting enteritis, diarrhoea and mesenteric lymphadenitis to severe systemic infections (Bottone, 1997). During infection, the bacteria first encounter the host mucosal surfaces overlying the intestinal cells and subsequently penetrate the epithelial cell layer through M cells into the lymphoid follicles, the Peyer's patches (Grutzkau et al., 1990). This process permits the bacteria to escape from host cell defences and provides a perfect opportunity to colonize and grow in regional lymph nodes and eventually in other tissues, such as kidney and liver, where they remain exclusively extracellular adherent and replicate outside the host cell (Grutzkau et al., 1990; Simonet et al., 1990).

Yersinia interaction with host cells is a critical first step in the pathogenesis of this organism and can be reproduced in tissue culture models (Bovallius and Nilsson, 1975). The addition of Yersinia to HEp-2 and HeLa cells in vitro results in tight adhesion followed by active uptake, with up to 20% of the added bacteria being internalized by the eukaryotic cells (Isberg et al., 1987). Uptake occurs via circumferential binding of the host cell about the surface of the bacterium with single individual microorganisms internalized into membrane-bound phagosomes (Devenish and Schiemann, 1981). Enteropathogenic Yersinia possess multiple mechanisms to enter mammalian cells (Miller and Falkow, 1988; Isberg, 1990; Yang and Isberg, 1993). It has been demonstrated that the outer membrane protein invasin is the most efficient factor that promotes cell entry to transfer enteropathogenic Yersinia through the epithelial layer (Pepe and Miller, 1993;Marra and Isberg, 1997). This protein is encoded by the chromosomal inv gene, which, when expressed in non-adherent Escherichia coli K-12 strains, is sufficient to enable this organism to bind and enter cultured cells efficiently (Isberg et al., 1987).

Functional analyses have shown that the N-terminal region of invasin is located in the bacterial outer membrane, whereas the C-terminal domain of the invasin protein is surface exposed and consists of five predominantly β-stranded subdomains (Leong et al., 1991; Hamburger et al., 1999). The two most extreme C-terminal domains are absolutely required for cell penetration and promote cell attachment and entry by binding to different members of the β1-integrin receptor family (Leong et al., 1990). Recent work has also shown that the second domain, extruding from the outer membrane, has the capacity to mediate intermolecular invasin–invasin interaction and significantly enhances the invasin-mediated uptake process (Dersch and Isberg, 1999; 2000).

Although the structure and function of the Y. pseudotuberculosis invasin has been characterized in detail, much less is known about its expression throughout the infection process. As invasion factors that promote initial uptake into host cells are of little or no use to the bacterium, except during the early stages of the infectious cycle, but render the bacterium more susceptible to the host immune responses, they are usually subject to tight and co-ordinate regulation (Bajaj et al., 1996; Porter and Dorman, 1997). As Yersinia internalization by host cells requires a relatively high amount of invasin in the outer membrane (Dersch and Isberg, 1999; 2000), a fine-tuned regulation of inv expression becomes a more compelling requirement for the transit of Yersinia through different host compartments. A large variety of different parameters, such as temperature, pH, osmolarity, oxygen, ion and nutrient content, varies between the various habitats outside and inside the host, and Yersinia could use these signals to sense which microenvironment they currently occupy and alter virulence gene expression accordingly. In fact, temperature is a key environmental cue for Yersinia to switch from a harmless, free-living microbe to an infectious agent that adheres to mammalian cells and secretes the virulence plasmid-encoded Yop proteins necessary for virulence (Cornelis and Wolf-Watz, 1997). The expression of all known invasion factors of Yersinia is thermally regulated in vitro (Straley and Perry, 1995). For instance, the ail and yadA genes, encoding outer membrane proteins that mediate adherence and low-level uptake into host cells, are expressed strongly at 37°C but repressed significantly at lower temperatures (Straley and Perry, 1995). In contrast, invasin expression in both Y. pseudotuberculosis and Y. enterocolitica occurs selectively at lower temperature (20–25°C) and is strongly repressed under 37°C (Isberg et al., 1988; Pepe et al., 1994). It is notable that a similar pattern of temperature regulation has also been observed for the enterotoxin Yst of Y. enterocolitica (Mikulskis et al., 1994). This observation is unexpected and would argue against their function in pathogenesis at a body temperature of 37°C. However, invasin has been detected in Y. enterocolitica isolated from the Peyer's patches of infected mice (Pepe et al., 1994), and it has been assumed that other environmental signals or nutrients that the organism encounters in the host might trigger invasin expression. In fact, experiments with Y. enterocolitica demonstrated that invasin expression can be significantly induced at 37°C when the pH of the medium is reduced below 7.0 (Pepe et al., 1994). Alternatively, other invasion factors, such as YadA or Ail, could participate in and stimulate the uptake process.

The control of invasive virulence factors, i.e. the inv/spa genes of Salmonella typhimurium or mxi/inv of Shigella flexneri, according to environmental signals is generally very complex and orchestrated by a cascade of differentially expressed regulatory factors (Bajaj et al., 1996; Porter and Dorman, 1997). Little is known about the underlying molecular mechanism and control factors, which respond to local environmental signals, involved in the regulation of the Y. pseudotuberculosis invasion genes. In this study, we have started to identify regulatory factors that control the expression of the Y. pseudotuberculosis inv gene in response to temperature and identified a transcriptional activator protein, homologous to the SlyA/Hor family (Ludwig et al., 1995; Thomson et al., 1997). Recently, Revell and Miller (2000) published a report on the SlyA homologue of Y. enterocolitica, RovA, which is also required for inv gene expression and virulence. The participation of the SlyA/Hor protein family in the regulation of stress response and pathogenesis prompted us to analyse the potential role of this regulator protein and some of its regulatory features in the environmental control of inv expression, which are presented here.


Construction and analysis of a chromosomal inv::phoA fusion in Y. pseudotuberculosis

To investigate invasin expression in response to environmental signals and identify regulatory factors that influence inv expression, an inv–phoA translational fusion was constructed, with codon 60 of invasin fused to the phoA gene, and transferred onto the Y. pseudotuberculosis chromosome (for details see Experimental procedures). A single recombination event yielded the merodiploid Y. pseudotuberculosis strain YP1, harbouring the inv wild-type copy and the inv–phoA fusion. In addition, a spontaneous segregant YP2 of this strain was isolated that had lost the integrated plasmid including the inv+ wild-type gene and only left the inv–phoA fusion at the inv locus of the Yersinia chromosome. The sequence and localization of the inv–phoA fusion in strains YP1 (inv+, inv–phoA) and YP2 (inv, inv–phoA) were verified by DNA sequencing and Southern blot analysis (data not shown). To ensure proper expression of the reporter gene fusion, the alkaline phosphatase activity of the strains was determined with regard to temperature and growth phase control and compared with the amount of wild-type invasin produced under the same environmental conditions (Fig. 1). As shown in Fig. 1C, the highest invasin expression of YP1 was found in stationary phase cells grown at 25°C, in which inv–phoA expression was about 10-fold induced compared with exponential cells grown at 37°C. This correlated with the level of invasin protein seen on Western blots, on which high amounts of the protein could only be visualized in cells grown at 25°C, whereas it was hardly detectable in exponential cells grown at 37°C (Fig. 1A). Low invasin expression was also found when only one condition (stationary phase or 25°C) was inducing. Under all conditions tested, regulation of inv–phoA expression was not effected by the presence or absence of the inv wild-type copy or the virulence plasmid pIB1 (data not shown). These results were all consistent with previous reports on temperature control of inv expression in Y. pseudotuberculosis (Isberg et al., 1988; Simonet and Falkow, 1992) and, furthermore, demonstrated that invasin is also under growth phase control.

Figure 1.

Effect of growth phase and temperature on the expression of inv and rovA. Y. pseudotuberculosis strains YP1 (inv–phoA) (A and C) and YP4 (rovA–lacZ) (B and D) grown in LB medium at 25°C and 37°C were harvested in the exponential phase (OD600 = 0.8) and stationary phase (OD600 = 4.5).

A and B. From equal cell numbers, whole-cell extracts were prepared, separated on a 10% SDS–PAGE and transferred on an Immobilon membrane. The wild-type invasin protein was visualized using monoclonal anti-invasin antibody mAb 3A2 (A), and the RovA–LacZ fusion protein was detected with a polyclonal antibody raised against β-galactosidase from E. coli (B). As noted previously by Isberg et al (1988), on Western blots using the mAb 3A2, the 108 kDa full-length invasion protein and several species lacking various amounts of the N-terminus are detected. The full-length invasin and the RovA–LacZ protein are marked by arrows; the sizes and positions of the molecular weight standard proteins are indicated on the left.

Alkaline phosphatase activity exhibited by strain YP1 (inv–phoA) (C) and β-galactosidase activity of YP4 (rovA–lacZ) (D) grown under the conditions indicated are shown. Alkaline phosphatase and β-galactosidase activity are expressed in µmol min−1 mg−1, and data represent the average ± SD from of at least three different experiments each performed in triplicate.

Identification of the inv transcriptional start site

To analyse invasin expression further and to determine the inv transcription start site, the complete nucleotide sequence of the 5′ non-coding region of the Y. pseudotuberculosis inv gene encoded by the chromosomal 1038 bp BamHI–SwaI fragment on pRI203 (Isberg et al., 1987) was determined. A close homologue to the flhE gene of Y. enterocolitica was identified 606 bp upstream of the invasin gene (Fig. 5A). This is the third gene of the flhBAE operon encoding components of the flagella apparatus. Thus, as in Y. enterocolitica, the inv gene of Y. pseudotuberculosis is located between the flagella genes flhBAE and flgNM (Fauconnier et al., 1997).

Figure 5.

Deletion analysis of the regulatory region required for inv expression.

A. Various lengths of the inv 5′ regulatory region were amplified by PCR and cloned upstream of the inv–phoA reporter gene in vector pPD264 using restriction endonucleases SwaI and BamHI. Restriction maps of the resulting plasmids are shown; the numbers given in brackets denote the size of the inv upstream region from the transcriptional start site. The location of the −10 and −35 promoter regions and the restriction sites used for the construction of the deletion plasmids are indicated.

B. Expression mediated by continuous deletions of the inv 5′ regulatory region upstream the inv–phoA reporter gene was analysed in E. coli CC118λpir/pGN7 (rovA+). Bacteria were grown overnight in LB medium at 37°C and 25°C. The alkaline phosphatase activities determined from the culture samples are given in µmol mg−1 min−1 and represent the mean ± SD of at least three independent experiments. CC118/λpir pACYC184 activity corresponds to the background level of the strains.

Primer extension analysis was used to identify the position of the 5′ end of the inv transcript. Total cellular RNA prepared from strain Y. pseudotuberculosis YPIII, grown overnight at 25°C and 37°C, was hybridized to an internal primer of the inv coding sequence. This primer was also used in a sequencing reaction to determine the start site of the primer extension product (Fig. 2A). In order to evaluate whether the activity of the inv–phoA fusion was directed by the same promoter predicted for the inv wild-type gene, primer extension analysis was also performed using total cellular RNA of Y. pseudotuberculosis YPIII(pPD264) grown at 25°C to exponential and stationary phase (Fig. 2B). An identical single signal was detected with both RNAs isolated from stationary phase cells grown at 25°C, whereas no band was seen with RNA isolated under conditions with either the temperature (37°C) or the growth phase (exponential phase) being repressive for invasin expression (Fig. 2A and B, lane 2). These results demonstrate that environmental control of inv expression occurs at the transcriptional level and confirm that the inv wild-type promoter is identical to the one used for invphoA transcription. The identified reverse transcription product corresponds to a single T residue 177 bp upstream of the inv translational start codon (Fig. 2A and B). Examination of the inv 5′ upstream sequence revealed the presence of a putative −35 (AGTCTG) and −10 (TATCCT) promoter sequence (Fig. 2C and D), correctly spaced by 17 bp.

Figure 2.

Identification and analysis of the inv promoter of Y. pseudotuberculosis.

A and B. Mapping of the inv and the inv–phoA transcription start site by primer extension analysis. Total RNA was prepared from strain YPIII (A) grown at 25°C (lane 1) or 37°C (lane 2) in LB medium or from strain YPIII pPD264 (B) grown at 25°C to stationary (lane 1) and exponential (lane 2) phase. For both experiments, a primer specific to the inv structural gene and 50 µg of inv or inv–phoA template RNA were used. The sequencing ladder obtained with the same primer is shown on the left. The arrow marks the identical +1 transcriptional start sites (A, lane 1; B, lane 1) 177 bp upstream from the inv start codon, and the identified −10 region is shown.

C. Comparison of the inv promoter regions of Y. pseudotuberculosis and Y. enterocolitica. The identified −35 and −10 promoter regions are boxed, and the homologous nucleotides are shadowed.

D. Nucleotide sequence of the inv promoter region. The inverted repeats IR1 and IR2 (potential RovA binding sites) are indicated with dark arrows, and the inverted repeat IR2, which includes one mismatch, is given in brackets. The +1 transcriptional start site as well as the sites used for the upstream deletion constructs are indicated by broken horizontal arrows. The numbers indicate the distance in bp from the transcriptional initiation site; the shadowed boxes show AT-rich motifs ≥ 12 bp. The −35 and −10 promoter regions determined according to the primer extension reaction are underlined (see A and B).

Identification of rovA, the gene encoding a transcriptional activator for invasin

Expression studies revealed that, in contrast to the temperature-dependent transcription of a plasmid-encoded inv–phoA fusion in Y. pseudotuberculosis YPIII, the same gene fusion was not expressed in E. coli K-12 grown at 25°C or 37°C (Fig. 3A). Furthermore, it has been shown that E. coli strains harbouring the inv gene on a multicopy vector showed no temperature dependence for cell penetration (Isberg et al., 1988). This suggested that the inv locus expressed in E. coli is no longer controlled by the same temperature-dependent regulatory circuit found in the parental Yersinia strain. Based on these observations, it seemed likely that an additional activating factor of Y. pseudotuberculosis was necessary for inv expression. To identify positive-acting components of invasin, a Y. pseudotuberculosis gene bank was introduced into E. coli strain CC118λpir harbouring the inv–phoA fusion plasmid pAK002. Approximately 10 000 clones were screened on agar plates containing X-P. Twenty dark blue colonies were identified and analysed further. After confirmation of the phenotype by retransformation of the plasmids into E. coli CC118λpir/pAK002 and restriction analysis, three independent plasmids (pAK011–013) were identified that induced a strong increase in inv–phoA expression in E. coli K-12 (20-fold) as well as in Y. pseudotuberculosis YP1 (fourfold) (Fig. 3B and C). The expression from the inv promoter in the presence of the complementing clones was significantly reduced by about 50–60% in both strains when cells were grown at 37°C (Fig. 3B and C).

Figure 3.

Expression of inv of Y. pseudotuberculosis in E. coli requires the presence of the RovA protein.

A. Y. pseudotuberculosis wild type and E. coli CC118λpir were transformed with the vector plasmid pACYC184, the inv–phoA fusion plasmid pPD264 or both pPD264 and pGN7 (rovA+) or pPD280 (rovA::Tn10).

B. E. coli strain CC118λpir carrying the inv–phoA fusion plasmid pAK002.

C. Y. pseudotuberculosis strain YP1 harbouring the chromosomal inv–phoA fusion was transformed with either the vector plasmid pACYC184 or the pACYC184-based gene bank plasmids pAK011, pAK012 and pAK013 carrying the rovA gene. The resulting strains were grown at 25°C or 37°C in LB medium, and alkaline phosphatase activity was determined from overnight cultures and is given in µmol min−1 mg−1. The data represent the average ± SD from three different experiments each performed in duplicate.

Sequence analysis revealed that the set of clones carried a single, common open reading frame (ORF), encoding a protein of 143 amino acids, which showed considerable amino acid identity to previously identified transcriptional activator proteins of the SlyA/Hor family of Enterobacteriaceae (Ludwig et al., 1995; Thomson et al., 1997). Because the identified transcriptional regulator exhibits its highest degree of identity (95%) to the SlyA homologue of Y. enterocolitica RovA/Hor, which has recently been shown to be involved in virulence (Bäumler and Hantke, 1992; Revell and Miller, 2000), we also refer to it as RovA.

The rovA coding sequence of Y. pseudotuberculosis starts with a TTG codon. This has also been shown for the slyA gene of E. coli and S. typhimurium, as well as for the rovA gene of Y. enterocolitica (Thomson et al., 1997). A consensus ribosome binding site is present 6–11 nucleotides upstream of the predicted start site of translation. No other ORF could be identified directly upstream of rovA in Y. pseudotuberculosis. However, 77 bp downstream of rovA, a short hypothetical ORF of 108 codons has been identified with no significant homology to known proteins. The function of this ORF is unclear, because it also overlaps with another ORF in the opposite orientation encoding the outer membrane lipoprotein Pcp/SlyB of 155 amino acids. This locus organization is identical to that found in Y. pestis and seems also to be conserved in Y. enterocolitica and other slyA+ bacteria (The Sanger Centre, http://www.sanger.ac.uk/Projects/Microbes/) (Bäumler and Hantke, 1992; Thomson et al., 1997).

To ensure that the RovA protein is solely responsible for temperature-dependent inv expression, the rovA gene was subcloned, and the resulting plasmid pGN7 was introduced into E. coli harbouring the inv–phoA fusion plasmid pPD264. As shown in Fig. 3A, no or only a little alkaline phosphatase activity was detected in E. coli harbouring the inv–phoA fusion plasmid pPD264. However, if both plasmid pPD264 and the rovA+ plasmid pGN7 were present in E. coli, inv expression was strongly induced at 25°C. The induction level was reduced to about 30% when bacteria were grown at 37°C (Fig. 3A). This temperature-dependent expression pattern of the inv–phoA fusion plasmid pPD264 in E. coli correlated strictly with that seen in the Y. pseudotuberculosis YPIII wild-type strain and allowed its use for further analysis of RovA-mediated inv expression in E. coli (see below).

Construction and characterization of a rovA mutant strain

In order to evaluate the activity of RovA on invasin synthesis and determine its influence on cell attachment, invasion and virulence, we constructed a rovA mutant strain. To do so, a mini-Tn10(Cm) insertion was isolated in codon 60 of the rovA gene encoded by plasmid pGN7 that totally abolished the expression of the inv–phoA fusion in E. coli(Fig. 3A). The insertion mutation was introduced into Y. pseudotuberculosis, and a stable recombinant strain, which was merodiploid (rovA+, rovA::Tn10) at the rovA locus, was selected. Subsequently, a rovA mutant strain was isolated, in which a second recombination event resulted in an allelic exchange that led to the loss of the plasmid including the rovA wild-type gene and only left the rovA::Tn10 insertion on the bacterial chromosome (for details, see Experimental procedures). The resulting rovA mutant strain YP3 was subsequently verified by polymerase chain reaction (PCR) and Southern blot analysis (data not shown).

To test whether the loss of the RovA protein had an effect on invasin synthesis, we compared the expression of the chromosomal inv–phoA fusion and the amount of the invasin protein in the Y. pseudotuberculosis wild type and the rovA mutant strain. To do this, we mated the suicide plasmid pAK002 harbouring the inv–phoA fusion into the rovA mutant strain YP3, generating YP10. Alkaline phosphatase activities from overnight cultures of YP1 (inv–phoA, rovA+) and YP10 (inv–phoA, rovA::Tn10) grown at different temperatures demonstrated that the absence of the transcriptional activator RovA led to a sixfold decrease in inv expression at 25°C and was similarly low when bacteria were grown at 25°C and 37°C (Fig. 4A). Western blot analysis with whole-cell extracts produced from equal amounts of Y. pseudotuberculosis YPIII (wild type) and YP3 (rovA::Tn10) overnight cultures showed a significantly lower amount of the invasin protein in the rovA mutant strain (Fig. 4B). As a third step in the analysis of the rovA mutant, we investigated whether the efficiency of Y. pseudotuberculosis adherence and entry into mammalian cells was dependent upon the presence of the rovA gene product. To do so, we performed an adhesion and invasion assay using HEp-2 cells. Figure 4C illustrates that the Y. pseudotuberculosis wild-type strain and the rovA mutant grown at 25°C interact equally well with HEp-2 cells (1.5%). However, rovA strain YP3 entered about sevenfold less efficiently than the YPIII wild-type strain. In comparison, the inv strain YP9 showed a 10-fold lower uptake rate than the parental strain. Similar adhesion and invasion properties of the Yersinia strains tested above were also found with other common cell lines, such as 3T3 and HeLa (data not shown).

Figure 4.

Analysis of a Y. pseudotuberculosis rovA mutant strain YP3.

A. Y. pseudotuberculosis strains YPIII, YP1 (inv–phoA) and YP10 (inv–phoA,rovA) were grown at 25°C and 37°C. Alkaline phosphatase activity from the strains was determined from overnight cultures and is given in µmol min−1 mg−1. The data represent the average ± SD from three different experiments each performed in duplicate.

B. Cell extracts were prepared from overnight cultures of Y. pseudotuberculosis strains YPIII (lane 2), YP9 (inv::Tn10) (lane 3) and YP3 (rovA::Tn10) (lane 4) separated on a 10% SDS–polyacrylamide gel and transferred onto an Immobilon membrane. The invasin protein was detected by the monoclonal antibody mAb 3A2 as described in Experimental procedures. A prestained protein marker was loaded in lane 1; the positions and corresponding molecular weights are indicated on the left. The invasin protein is shown by an arrow.

C. Cell binding and entry efficiency of E. coli K-12 and Y. pseudotuberculosis strains YPIII, YP9 (inv) and YP3 (rovA). About 5 × 106 bacteria of an overnight culture grown at 25°C were used to challenge 105 HEp-2 cells and incubated for 1 h at 20°C to monitor cell adhesion or at 37°C to determine the invasion efficiency. The total number of surface-bound bacteria was determined by plating the bacteria after separation from cells, and the number of intracellular bacteria was determined by the gentamicin protection assay as described in Experimental procedures. Each bar represents the average of three independent experiments, and the superimposed error bars indicate ± SD.

Deletion analysis of the upstream sequence required for inv expression

Sequence analysis revealed several direct repeats and short palindromic sequences in the 5′ regulatory region of inv that might be involved in the RovA-mediated activation of inv transcription. We have shown previously that the inv–phoA fusion encoded by pPD264 can be used to monitor RovA-dependent inv expression in E. coli under different environmental conditions (see Fig. 3A). In order to define the upstream sequence that is required for invasin expression, a series of translation fusions (based on pPD264) with progressively shorter sequences of the inv regulatory region were constructed, as described in Experimental procedures(Fig. 5A). The resulting plasmids were transformed into E. coli strain CC118λpir harbouring the rovA+ plasmid pGN7, and the alkaline phosphatase activity of the resulting strains, grown at 25°C and 37°C, was determined (Fig. 5B). The RovA-mediated activation of inv expression was not affected by promoter deletions upstream from position −207. As shown by the alkaline phosphatase activity resulting from the fusions encoded by pGN1–3, equal amounts of invasin are expressed, with a threefold higher concentration at 25°C versus 37°C (Fig. 5B). However, the inv–phoA fusion encoded by pGN8, extending from position −90, was clearly reduced and retained only 70–80% of RovA-mediated inv activation compared with pGN1 including the entire inv regulatory region. Fusions up to the positions −50 (pGN4) and −31 (pGN5), including either the −10 or both promoter regions, produced only about 30–40% of the alkaline phosphatase activity obtained with the full-length promoter region on pPD264. Furthermore, inv expression of both fusions was very similar at 25°C and 37°C, indicating that temperature control was lost in these fusion constructs. As expected, the construct including the region downstream of position +23 (pGN6) without the inv promoter did not exhibit significant inv–phoA expression in E. coli, comparable with the empty vector control. This is equivalent to the transcriptional activity of all the fusion constructs in the absence of the rovA regulator gene, which does not exceed 0.16 µg mg−1 min−1 (data not shown). Taken together, all cis-acting DNA elements required for optimal temperature-regulated and RovA-mediated invasin expression must be located downstream of position −207 and upstream of position −50.

Using E. coli, the influence of other specific regulatory factors of Yersinia that might also interact with the inv promoter was excluded. To investigate whether additional regulatory components might be involved, a selection of the promoter deletion constructs was also introduced into Y. pseudotuberculosis strains YPIII and YP3 (rovA), and inv–phoA expression was analysed as described for E. coli. The introduction of the fusion plasmid into Y. pseudotuberculosis yielded the identical expression pattern to that in E. coli with or without rovA (data not shown), indicating that RovA-mediated inv expression is identical in Y. pseudotuberculosis and E. coli.

Overproduction and purification of RovA

In order to provide physical evidence for the interaction of RovA with the inv promoter region, soluble purified RovA protein was needed. For this purpose, a C-terminal His-tagged RovA hybrid protein (RovA-6His) was synthesized in E. coli K-12. RovA-6His conferred invasin expression, indicating that the hybrid protein was active. After induction of rovA expression, a polypeptide with a molecular weight of 18 kDa was detected on SDS gels and Western blots using a His-tag-specific monoclonal antibody (data not shown), which is consistent with the molecular weight of RovA (18.6 kDa) deduced from its DNA sequence. The recombinant RovA-6His protein was soluble (data not shown) and was purified to > 95% homogeneity by affinity chromatography using Ni-NTA agarose (Fig. 6A; for details, see Experimental procedures).

Figure 6.

A. Purification of RovA. Strain XL1blue (pGN10, rovA+) was grown in the presence of the inducer IPTG in LB medium, cell extracts were prepared, and the RovA protein was purified by affinity chromatography using Ni-NTA agarose. Purified RovA protein was separated on a 15% SDS–polyacrylamide gel (lane 2); lane 1, marker proteins (Mr indicated on the left). The position of the RovA protein is indicated by an arrow.

B–D. Gel retardation experiments using purified RovA protein.

B. Individual DNA fragments b and c (1 µg of each) comprising different portions of the inv upstream region (b, −207 to +223; c, −90 to +223) were incubated without protein (lane 1) or with increasing amounts of purified RovA protein (1.0, 1.1, 1.2, 1.3 and 1.5 µg; lanes 2–6).

C. Competitive gel shift assay with inv promoter fragments a, c (a, −426 to +223; c, −90 to +223) and the control fragment d without the inv promoter (d, +23 to +223) were incubated without protein (lane 1), with 10 µg of BSA (lane 2) or with increasing amounts of purified RovA protein (0.8, 0.9, 1.14, 1.3, 1.45 and 1.6 µg; lanes 4–9).

D. Competitive gel shift assay with the inv promoter fragment c (c, −90 to +223) and the control fragment e encompassing the csiD promoter (−127 to +109) of E. coli K-12 were incubated without protein (lane 1) or with increasing amounts of purified RovA protein (1.0, 1.1, 1.2, 1.3 and 1.5 µg; lanes 2–6). The DNA–RovA complexes were separated on a 4% polyacrylamide gel, a molecular weight standard M (100 bp ladder) was loaded, and the corresponding molecular weights are indicated on the left. The position of the inv promoter and the inv and csiD control fragments are indicated; the higher molecular weight complexes are shown by arrows.

Interaction of Rov with the inv promoter region

The DNA-binding capacity of purified RovA for the inv promoter region was investigated by gel mobility shift assays. PCR fragments encompassing different portions of the inv promoter region, equivalent to the control region in pGN1–6 and pGN8 (Fig. 5A), were incubated individually with increasing concentrations of purified RovA and assayed for protein–DNA complex formation. Two higher molecular weight complexes were found with the inv promoter fragments (a and b) extending up to position −207 or further (Fig. 6B and C), indicating that at least two RovA molecules interact with the entire inv regulatory region. A single RovA–DNA complex was detected when the inv regulatory region was shortened to position −90 (fragment c; Fig. 6C and D). In contrast, no complex formation was found with the fragments encompassing inv sequences downstream of position −50, equivalent to pGN4, 5 and 6 (data not shown). To compare the DNA-binding capacity and specificity of RovA with individual DNA fragments, we performed competitive gel retardation assays with fragments of the inv promoter (Fig. 6C) and a foreign control fragment encompassing the promoter region of the csiD gene of E. coli(Fig. 6D). Identical RovA–DNA complex formation was found with the inv promoter fragments a and c, equivalent to pGN1 and pGN3, whereas no binding was seen with the csiD or the inv fragment without promoter, equivalent to pGN6.

From structural analysis of the MarR repressor protein, it is known that the dimeric form of this protein interacts with the DNA and recognizes short palindromic sequences separated by 3–4 bp (Alekshun et al., 2000). Because RovA has a similar size and exhibits significant homology to MarR (29%), one can assume that its recognition sites might be somewhat similar. In fact, two similar 5 bp palindromic sequences were found on the large inv promoter fragments preferentially shifted by RovA (Fig. 2D). The sequence ATATTN3AATAT, designated IR1, is located at −62 to −50 relative to the inv start codon. The second motif, IR2, located at position −172 to −160, harbours one mismatch to IR1 and could thus be a second lower affinity binding site for RovA. Interaction of RovA with IR1 and IR2 would explain maximal inv induction seen with the deletion constructs harbouring inv promoter fragments extending to position −207 (Fig. 5). The smaller fragment c, in which the IR2 motif is completely deleted, but carrying the IR1 site, is still able to bind RovA preferentially; however, slightly higher protein concentrations seemed to be needed to achieve comparable complex formation (Fig. 6B and C). This is in full agreement with our previous results, in which we showed that this promoter fragment, cloned on pGN8, was still able to induce inv transcription in E. coli to 70–80% (Fig. 5B). Our data strongly suggest that, at low concentrations (1 µg of RovA µg−1 DNA), RovA exhibits a certain preference for the inv regulatory region including positions −207 to −50. However, it is also notable that an increase in the protein concentrations (> 3 µg of RovA µg−1 DNA) led to the binding of RovA to all DNA fragments tested and to the appearance of higher molecular weight DNA complexes in vitro.

Effect of temperature on rovA expression

As RovA encodes a transcriptional activator and is essential for high-level expression of invasin, the regulation of invasin might be mediated by the regulation of rovA expression. To test this hypothesis, we first analysed whether rovA expression is also regulated by temperature and growth phase, as shown previously for inv(Fig. 1A and C). To facilitate the analysis of rovA expression, we constructed a rovA–lacZ fusion on a mobilizable suicide vector (pPD281) that carries the regulatory region of rovA extending −325 bp upstream of the rovA initiation codon, with codon 129 being fused in frame to a promoterless E. coli lacZ gene. Using the plasmid mating strategy into Y. pseudotuberculosis described earlier, we were able to construct a merodiploid rovA+, rovA–lacZ fusion strain YP4 (for details, see Experimental procedures). As a first step in rovA regulation, YP4 (rovA–lacZ) was grown in LB at 25°C and 37°C to exponential or stationary phase and compared with YP1 (invphoA) grown under identical environmental conditions. As judged by Western blot analysis and β-galactosidase assays (Fig. 1B and D), maximal expression of the rovA–lacZ fusion was observed in stationary phase cells grown at 25°C, whereas significantly lower levels of the fusion protein were detected if bacteria were grown exponentially and/or at 37°C. Thus, the expression pattern of rovA is fully consistent with that observed for inv expression (Fig. 1A and C), suggesting that environmental control of invasin regulation is mediated through RovA. To investigate further the influence of rovA induction on inv transcription in Y. pseudotuberculosis, we followed the expression of the inv–phoA and rovA–lacZ fusions along the bacterial growth curve. To do so, Yersinia strains YP1 (inv–phoA) and YP4 (rovA–lacZ) were grown overnight in LB and diluted 500-fold into fresh LB medium. Growth, β-galactosidase and alkaline phosphatase activities were determined continuously and simultaneously from culture samples grown at 25°C and 37°C taken every hour. We were able to show that rovA expression during growth at 25°C was activated in mid-log phase and increased continuously throughout the growth curve, resulting in maximal rovA expression during stationary phase (Fig. 7A). rovA expression also increased during growth at 37°C, but remained fivefold lower compared with bacteria grown at 25°C. Notably, inv expression in 25°C-grown cells exhibited a similar but delayed induction curve that began to increase dramatically as the bacteria entered early stationary phase and reached its maximum in late stationary phase. These data are in full agreement with previous results in this study (Fig. 1) and indicate that a certain amount of presynthesized RovA protein is needed for efficient activation of inv expression.

Figure 7.

Effect of growth phase and osmolarity on the expression of inv and rovA.

A. Overnight cultures of strains YP1 (invphoA) and YP4 (rovA–lacZ) were diluted 1:500 in fresh LB and grown at 25°C or 37°C as indicated. The growth of the cultures (closed symbols) was
monitored continuously at OD600. Alkaline phosphatase activity
and β-galactosidase activity (open symbols) of culture samples taken every hour are plotted for comparison. The values are the mean of three independent experiments.

B. Strains YP1 and YP4 were grown at 25°C or 37°C in LB without or with increasing concentrations of NaCl (100–600 mM). Alkaline phosphatase activity and β-galactosidase activity are plotted in µmol min−1 mg−1 observed for each fusion strain and represent the mean of at least three independent experiments.

Expression of rovA and inv follows the same pattern of regulation in response to growth phase, growth media, pH, osmolarity and ion concentration

The results described above prompted us to examine whether rovA expression was also modulated by other environmental cues, such as ion and nutrient concentration, osmolarity and pH. These parameters change when Yersinia enters a host or a different host compartment and have been shown previously to regulate the expression of invasin genes of other related pathogens (Pepe et al., 1994;Bajaj et al., 1996; Martinez-Laguna et al., 1999).


Changes in osmolarity have been found to influence the transcription of virulence genes in several genera of bacteria, including E. coli, Salmonella and Vibrio, and it is known that optimal gene transcription occurs at different osmolarities depending on the specific infection site (Mekalanos, 1992). To analyse the influence of osmolarity on rovA and inv expression of Y. pseudotuberculosis, strains YP1 and YP4 were grown in LBON (LB without NaCl) supplemented with 0 and 100–600 mM NaCl at concentrations that did not significantly alter the cell density of the overnight cultures. As shown in Fig. 7B, inv–phoA and rovA–lacZ were maximally expressed at 100–200 mM salt. This corresponds well with the physiological osmolarity in the gut (0.11 M), particularly in the lumen of the ileum (Sleisenger, 1981;Hallback et al., 1991). A further increase in osmolarity led to a gradual decrease in inv and rovA expression. In contrast, the expression of both genes was repressed (0,04 µmol mg−1 min−1) and not significantly altered by osmolarity when the bacteria were grown at 37°C (Fig. 7B; for rovA, data not shown). To ensure that this effect was caused by an increase in osmolarity and not an increase in the NaCl concentration in the medium, NaCl was replaced by KCl or arabinose, a sugar that cannot be used by Y. pseudotuberculosis as a carbon source. Similarly, the expression of inv and rovA was also maximal at a final concentration of 100–200 mM of the added molecules, although the addition of the saccharide only led to a twofold induction of both genes (data not shown).


It has also been demonstrated that the expression of the inv gene of Y. enterocolitica is strongly increased at 37°C when the bacteria were grown under low pH (Pepe et al., 1994). Thus, in Y. enterocolitica, invasin expression seems to be induced at 37°C in the host, when the bacteria encounter the stomach during the first phase of infection or induce a local acidity after the colonization of the terminal ileum (Sleisenger, 1981). In contrast to the inv gene of Y. enterocolitica, invasin as well as rovA expression of Y. pseudotuberculosis were not significantly altered by changes in the pH (5.5–8.5) when YP1 and YP4 were grown at 37°C (Fig. 8A). This result was confirmed independently with different LB media buffered in the same range from pH 5.5 to 8.5 using different buffer substances, such as MES, PIPES and MOPS. In all experiments, the pH of the overnight cultures did not change more than 0.2 units. Figure 8A also shows that inv and rovA expression at 25°C remained elevated at pH 7–8.5, equivalent to the pH of the ileum, caecum and colon, but decreased continuously when the pH dropped below pH 7. At pH 5.5, inv and rovA expression of Y. pseudotuberculosis was reduced about five- to sixfold compared with growth at pH 7.

Figure 8.

Effect of pH and growth phase on the expression of inv and rovA.

A. Strains YP1 (invphoA) and YP4 (rovA–lacZ) were grown overnight at 25°C or 37°C in buffered LB medium (with 0.1 M MES or MOPS). Alkaline phosphatase and β-galactosidase activity are plotted in µmol min−1 mg−1 for comparison and represent the average ± SD of at least three different experiments.

B. Effect of growth media on inv and rovA expression. Strains YP1 and YP4 were grown at 25°C in LB and minimal media MOPS, DMEM, RPMI, M9 and M63. Alkaline phosphatase activity and β-galactosidase activity were determined from cultures grown to stationary phase and are plotted in µmol min−1 mg−1 for comparison. The data represent the mean of at least three independent experiments.

Nutrient content

The effect of variations in the nutrient content on the expression of rovA–lacZ was analysed in parallel with the inv–phoA fusion by growing the fusion strains YP1 and YP4 in different growth media at 25°C and 37°C to late stationary phase. LB mimics the nutrient-rich environment of the host intestine, whereas minimal media M9, MMA and M63 mimic the poor external environment. As growth in tissue culture media is known to stimulate the expression of virulence genes (Martinez-Laguna et al., 1999), activation of inv and rovA expression in RPMI and DMEM was also analysed. Compared with growth in LB medium, rovA–lacZ at 25°C showed less activity in the minimal media (12% activity in MMA, M9 and M63 and 30% activity in RPMI and DMEM). This was proportionally similar to the reduction seen for inv–phoA under the same conditions (Fig. 8B). This is in contrast to the results with the inv gene of Y. enterocolitica, for which no effect of the growth medium was reported (Pepe et al., 1994).

Various micronutrients (such as Ca2+, Fe2+, Mg2+, Mn2+ and ammonia) are known to influence the expression of virulence genes in a large number of bacterial pathogens (Mekalanos, 1992). In addition, environmentally controlled virulence genes of Yersinia and other pathogens are often affected by DNA supercoiling and require a certain degree of DNA superhelicity or AT-rich curved DNA sequences for their expression (Dorman et al., 1990; Mekalanos, 1992; Mikulskis et al., 1994; Rohde et al., 1994; 1999). However, neither inv nor rovA expression was affected by the addition to the growth media of different amounts of micronutrients or topoisomerase inhibitors (data not shown). Although the inv regulatory region contains several poly(AT) stretches (Fig. 2C), other typical characteristics of bent DNA, such as differences in the electrophoretic mobility at 4°C and 60°C of the inv promoter fragment, could not be detected (data not shown).

Autoregulation of rovA

Transcriptional activator proteins of virulence genes are often involved in the synthesis of a large number of genes, and they usually tightly control their own synthesis (Brown and Taylor, 1995; Martinez-Laguna et al., 1999). To assess whether rovA is subject to an autoregulatory mechanism, the rovA–lacZ fusion encoded by the plasmid pPD281 was introduced into the chromosome of the Y. pseudotuberculosis rovA mutant strain YP3, yielding YP6 (rovA–lacZ, rovA::Tn10), as described in Experimental procedures. The β-galactosidase activity of YP6, compared with the equivalent rovA+ strain YP4 (rovA–lacZ), both grown at 25°C, showed that the elimination of rovA reduced the expression of rovA–lacZ to 20% of the wild-type activity (Fig. 9A). The expression level of rovA in a rovA background at 37°C is almost identical to that in the wild-type strain. This result indicates that RovA activates its own expression depending on the environmental conditions sensed by the organism. However, it is unclear whether RovA regulates its own expression directly or indirectly. To confirm further that RovA promotes its own synthesis, rovA+ plasmid pGN7 was transformed into YP6. The β-galactosidase activity of this strain showed that, when rovA was placed in trans, it restored activation of rovA as well as inv expression (Figs 3 and 9A).

Figure 9.

A. Positive autoregulation of rovA expression. Strains YP4 (rovAlacZ), YP6 (rovA–lacZ, rovA::Tn10) and YP6 harbouring the rovA+ plasmid pGN7 were grown in LB medium at 25°C and 37°C. β-Galactosidase activity from overnight cultures was determined and plotted in µmol min−1 mg−1 for comparison.

B. Post-transcriptional regulation of inv expression. E. coli strain CC118λpir pAK002 (inv–phoA) was transformed with pHSG575 (plac) or pPD293 (plac::rovA). The resulting strains were grown in LB or MMA supplemented with 0.2% glycerol or lactose at 25°C or 37°C. Alkaline phosphatase activity from overnight (OD600 = 4.5) and exponential cultures (OD600 = 0.8) was determined and plotted in µmol min−1 mg−1 for comparison. The data represent the average ± SD from at least three different experiments each performed in duplicate.

RovA expression is regulated at the post-transcriptional level

Under all the conditions tested in this study, rovA expression follows exactly the same pattern of regulation as inv expression. This provides evidence that inv expression in response to environmental signals is mediated by regulation of RovA. This could be established by changes in rovA transcription induced by certain, if not all, environmental conditions. Alternatively, modulation of RovA activity by environmental signals could account for the co-ordinate regulation of rovA and inv. In this case, inv and rovA expression should be independent of any regulation of the rovA promoter, and invasin should be environmentally controlled even when the rovA gene is expressed from a foreign inducible promoter. To test this hypothesis, E. coli strain CC118λpir/pAK002 (inv–phoA) was transformed with a plasmid pPD293, which harbours the rovA gene under the control of the inducible lac promoter. Invasin expression was monitored under different environmental conditions, as described previously. It was found that invasin expression controlled by the lac promoter was significantly reduced at 37°C, in exponential phase and/or in minimal medium with or without lactose (Fig. 9B). Similar results were observed for rovA–lacZ and invphoA expression in Y. pseudotuberculosis when plac ::rovA was introduced into the rovA mutant strains YP6 and YP10 respectively (data not shown). These results argue that temperature, nutrient and growth phase control of rovA expression is independent of the rovA promoter and suggest a post-transcriptional mechanism of environmental control of RovA synthesis.


Enteropathogenic Yersinia as well as other pathogens are able to sense a variety of environmental stimuli. They use this information to determine whether they are outside or within a host and then regulate virulence gene expression accordingly in order to use their resources more efficiently and prevent host immune responses (Mekalanos, 1992; Straley and Perry, 1995). As shown in this and previous studies (Isberg et al., 1988; Simonet and Falkow, 1992) invasin, the primary invasive factor of Y. pseudotuberculosis, is an environmentally responsive virulence determinant, whose production is modulated by a variety of external signals, such as temperature, osmolarity, growth phase, growth medium and pH. To analyse the molecular mechanism that mediates the environmental control of inv expression in Y. pseudotuberculosis, we made a first attempt to identify regulatory factors and characterize their function in the inv regulatory process. We found that inv regulation occurs at the transcriptional level and requires a transcriptional activator protein, highly homologous to the SlyA/Hor family (Ludwig et al., 1995; Thomson et al., 1997). The regulatory gene was identified by a genetic complementation strategy that activates the expression and restores temperature regulation of an unexpressed inv–phoA fusion in E. coli, suggesting that the activator protein acts directly on the inv promoter and that no additional factors seem to be necessary for this interaction.

The SlyA protein was first identified in S. typhimurium, in which it plays an important role in virulence in mice and in the survival of the bacteria in the intracellular environment of host macrophages (Libby et al., 1994;Buchmeier et al., 1997). In Salmonella infections, SlyA regulates the expression of a number of proteins during stationary phase and upon phagocytosis by macrophages (Daniels et al., 1996;Buchmeier et al., 1997). Its expression is required for resistance to oxidative stress and the destruction of M cells, but not for invasion or colonization of the murine small intestine (Daniels et al., 1996). Studies in E. coli have also demonstrated that SlyA of S. typhimurium can activate the transcription of the cryptic clyA gene encoding cytolysin A, a member of the RTX toxin family that confers a haemolytic phenotype (Ludwig et al., 1995; Oscarsson et al., 1996;Fernandez et al., 1998). slyA homologues have also been detected in other pathogenic bacteria, such as Shigella, Citrobacter, Serratia, Enterobacter and Erwinia, illustrating their wide distribution in the bacterial world. A homologous gene has also been identified recently in the mutS–rpoS locus located in a variable genomic region at 61 min on the E. coli chromosome of pathogenic E. coli, such as enterohaemorrhagic E. coli (EHEC) and enteropathogenic E. coli (EPEC), but was not found in E. coli K-12 or members of the ECOR group A (Herbelin et al., 2000). Preliminary data showed that inv–phoA expression was significantly induced in E. coli strains derived from the EPEC group (P. Dersch, unpublished results), suggesting that rovA function can be complemented by the slyA gene of E. coli. Furthermore, the absence of the slyA gene in E. coli K-12 strains explains why a plasmid-encoded inv–phoA fusion in our study was not expressed unless the Y. pseudotuberculosis slyA homologue was introduced. A function for SlyA in pathogenesis was also suggested by results from a recent study with the slyA homologue of Y. enterocolitica, named rovA (Revell and Miller, 2000). This work showed that rovA is required for invasion and efficient colonization of the Peyer's patches, the mesenterial lymph nodes and the spleen of infected mice. A rovA mutant strain of Y. enterocolitica showed a 70-fold increase in the LD50 compared with wild type and was also significantly attenuated in the mouse infection model. As the Y. pseudotuberculosis transcriptional regulator found in this study exhibits 95% amino acid identity to the RovA protein of Y. enterocolitica, we also refer to it as RovA.

RovA is also distantly related to a broad family of bacterial regulatory proteins controlling diverse functions of bacterial physiology, such as repression of microcin production, multiple antibiotic resistance and growth repression in E. coli. The family includes the repressor proteins MrpA, MarR, HpcR and Prs from E. coli, Hpr from Bacillus subtilis and PecS from Erwinia chrysanthemi. All these regulators govern functions crucial for survival, i.e. inactivation of toxic compounds, acquisition of resistant phenotypes and cytotoxicity to host cells (Thomson et al., 1997). The mechanism of regulation, however, is obviously different, as these proteins act as repressor proteins, whereas RovA is a transcriptional activator. Presumably, the homologies at the amino acid level indicate a similar domain structure with the ability to bind DNA.

All these proteins seem to regulate multiple genes and operons. In order to analyse the function of RovA in Y. pseudotuberculosis, we constructed a rovA mutant strain. As invasin expression was significantly reduced in the rovA strain as well as in E. coli K-12 (slyA) (Fig. 3), inv expression seems to be fully dependent on a functional RovA protein. Moreover, we showed that RovA has no influence on cell adhesion, but plays an important role in invasin-mediated uptake of Y. pseudotuberculosis into mammalian cells. Presumably, cell adhesion and the remaining inefficient cell entry of a Y. pseudotuberculosis rovA mutant is mediated by the YadA protein, as shown for an inv mutant strain (Yang and Isberg, 1993;Marra and Isberg, 1997). Furthermore, previous quantification of invasin molecules necessary for uptake and adhesion would predict that a strong reduction in the amount of invasin, as seen in the rovA mutant (Fig. 3B), would dramatically reduce the entry efficiency but would not significantly influence cell adhesion (Dersch and Isberg, 1999; 2000).

In order to define the molecular mechanisms involved in the environmental regulation of inv expression, we identified the inv promoter and investigated the interaction of the RovA protein with the inv regulatory region. The −10 region of the inv promoter is homologous to σ70-dependent E. coli promoters, whereas the potential −35 region exhibits no matches with the conventional −35 promoter sequence and also has no homology to the −35 region of the Y. enterocolitica inv gene (Fig. 2C; Pepe et al., 1994). Although homologous sequences were identified in both the Y. enterocolitica and the Y. pseudotuberculosis inv promoter region (Fig. 2C), a deletion analysis revealed that nucleotides up to position −50, including the promoter and the homologous sequences, exhibited the same low promoter activity as a construct that harbours only the −10 region. These observations emphasize that a particular −35 region does not seem to be necessary for RovA-driven inv transcription, and the conserved promoter regions may contribute operator sequences for other common, not yet identified regulatory factors.

Our deletion analysis of the inv regulatory region also revealed that all cis-acting regulatory DNA elements necessary for maximal temperature and RovA-mediated inv expression are located between positions −207 and −50 in the inv regulatory region. Competitive DNA bandshift assays with purified RovA protein were consistent with this observation. In these experiments, RovA exhibited its highest affinity for DNA fragments, including the −207 to −50 inv promoter region, and led to the formation of two RovA–DNA complexes. The presence of two different RovA–DNA complexes could result from the interaction of multimeric RovA complexes caused by RovA–RovA interaction or might indicate two different RovA binding sites in the inv regulatory region. Two similar 5 bp palindromic sequences, IR1 and IR2 (Fig. 2D), which differ by one nucleotide, were identified in the inv promoter segment and constitute two potential RovA binding sites that may differ slightly in their affinity for RovA. Presumably, sequential binding of RovA to both sites is necessary to mediate optimal, temperature-regulated inv activation, as only inv promoter fragments including both potential RovA binding sites confer maximal inv activation under low temperature conditions, whereas a fragment that harbours only IR1 and allows the formation of only one RovA–DNA complex exhibits a 20–30% reduced promoter activity (Fig. 5B). Interestingly, both inverted repeats are located in AT-rich regions, indicating that either the binding sites have to be flexible for RovA–DNA complex formation or AT-rich primary structures might generally serve as preferential RovA binding sites. However, to prove sequence-specific binding of RovA clearly and define the RovA binding sites, future DNA footprinting experiments and the analysis of point mutations that disrupt RovA binding are needed.

Studies on transcriptional control of invasin within Y. pseudotuberculosis in this study have revealed similar but distinct regulatory mechanisms to invasin in the related enteric pathogen Y. enterocolitica. In both Yersinia strains, inv gene expression is maximally induced by RovA in late stationary phase at a growth temperature of 25°C (Fig. 1;Pepe et al., 1994; Revell and Miller, 2000). These inducing conditions for invasin expression are found in the external environment of the bacteria rather than inside the host. Hence, an increased synthesis of invasin outside the host could prime enteropathogenic Yersinia for infection, as a sudden uptake of the organisms into the intestinal tract of the host is usually followed shortly by their penetration into M cells (Pepe et al., 1994; Marra and Isberg, 1997). In contrast, regulation by nutrients and the pH of the culture medium seem to be quite different in both Yersinia strains. In Y. enterocolitica, it has been shown that inv expression is not affected by the growth medium, and inv transcription is significantly induced at 37°C if the pH of the medium is reduced below 7.0 (Pepe et al., 1994). In Y. pseudotuberculosis, however, no such induction was found (Fig. 8A). Thus, although the identical regulatory protein is used for inv expression, a differential regulation pattern can be achieved in the related microorganisms, which might enable them to adapt individually to their preferred colonization sites.

The differences in the regulation pattern of invasin between the two Yersinia species might have been adapted by the organisms as a result of the different properties of the individual invasin molecules. As shown previously, the invasin protein of Y. pseudotuberculosis can form multimers in the bacterial outer membrane that dramatically increase the efficiency of the uptake process. In contrast, the multimerization domain of the invasin protein is naturally deleted in Y. enterocolitica; hence, invasinent is not able to form multimers in the bacterial outer membrane and promotes an invasion process that is much less efficient than that observed with the Y. pseudotuberculosis homologue. However, invasion efficiency is highly dependent on the amount of invasin in the outer membrane, and this defect can be somewhat overcome by the overproduction of invasinent (Dersch and Isberg, 2000). Therefore, an elevated expression of invasinent at 37°C, induced under the low pH conditions found in the stomach and some parts of the small intestine (Sleisenger, 1981), may be advantageous for the invasion process of Y. enterocolitica in its mammalian host.

In an attempt to define the mechanism involved in environmental and RovA-mediated regulation of inv expression, we found that the regulation of invasin by environmental signals is mediated through RovA. Comparative regulatory analysis of inv and rovA expression in our study revealed the identical regulation pattern of both genes in response to all environmental stimuli tested. Furthermore, both genes are expressed throughout the bacterial growth curve, with the rovA gene being maximally activated in early stationary phase, whereas inv gene expression was maximally induced in late stationary phase. According to these data, a certain number of RovA molecules seems to be necessary to activate directly inv expression and probably other factors involved in virulence. Differences in the regulation pattern of inv expression in Y. enterocolitica and Y. pseudotuberculosis can thus be caused by differences in either rovA induction or other regulatory factors that influence the interaction of RovA with the inv regulatory region under certain environmental conditions.

In this report, we also found that, in addition to the inv gene, RovA activates the expression of its own gene. Positive autoregulation has not yet been described for members of the SlyA family, but is a common feature of other transcriptional activators of virulence genes, such as BfpT, BvgA and ToxT of the AraC activator family (Roy et al., 1990;Brown and Taylor, 1995; Martinez-Laguna et al., 1999). In our experiments, only a low basal expression was detected in the absence of RovA or in the absence of the inducing environmental signal (Fig. 9). Presumably, low levels of RovA are maintained during non-inducing conditions, so that the activator remains readily available to turn on the regulatory circuit under inducing conditions. A complex set of physiological cues seems to be sensed by the organism, which responds by regulating RovA activity and thus rovA expression. This raises the question of what allows the upregulation upon the appearance of the appropriate environmental signals. The identity of the sensor and regulator proteins involved in this process are still unknown and will be the subject of further investigations.

Experimental procedures

Bacterial strains, cell culture, media and growth conditions

The strains used in this study are listed in Table 1. Overnight cultures of E. coli were grown routinely at 37°C; Yersinia strains were grown at 25°C in LB (Luria–Bertani) broth if not indicated otherwise. To analyse the effect of different growth media, LB broth, M9, MMA, M63 (Sambrook et al., 1989), DMEM and RPMI (Gibco BRL) containing glucose (0.4%) were used to culture Y. pseudotuberculosis at 25°C and 37°C as indicated. To monitor the influence of osmolarity, the strains were grown in LB without NaCl (LBON) and modified by the addition of 100, 200, 300, 400, 500 and 600 mM NaCl, KCl or arabinose. Variations in pH were obtained by buffering LB with MOPS, MES and PIPES at 0.1 M and then adjusting the pH to different values by the addition of NaOH or HCl.

Table 1. Bacterial strains and plasmids.
Strains, plasmidsDescriptionSource or reference
  1. a . inv–phoA fusion protein expressed from the inv promoter; the numbers indicate the codon of inv pstb fused to TnphoA.

  2. b . The numbers indicate the codon of rovA pstb in which the Tn10 has been inserted.

  3. c . The numbers indicate the codon of rovA pstb fused to lacZ.

Bacterial strains
 E. coli K-12
  XL1blueF′::Tn10 proA+B+lacIqΔ(lacZ)M15/recA1 endA1 gyrA96 (Nalr) thi hsdR17 (rK mK+) supE44 relA1 lac Bullock et al. (1987)
  CC118λpirFΔ(ara-leu)7697 Δ(lacZ)74 Δ(phoA)20 araD139 galE galK thi rpsE rpoB arfEamrecA1, λpir Manoil and Beckwith (1986)
  S17-1λpir recA thi pro hsdR M+ (RP4-2 Tc::Mu-Km::Tn7), λpir Simon et al. (1983)
 Y. pseudotuberculosis
  YPIIIpIB1, wild type Bolin et al. (1982)
  YP1pIB1, inv::TnphoA(60)a, KanRThis study
  YP2pIB1, inv::TnphoA(60)ainv, KanRThis study
  YP3pIB1, rovA::Tn10(60)b, CmlRThis study
  YP4pIB1, rovA-lacZ(129)c, AmpRThis study
  YP6pIB1, rovA-lacZ(129)c, rovA::Tn10(60)b, AmpR, CmlRThis study
  YP9pIB1, inv::Tn10, TetRThis study
  YP10pIB1, inv::TnphoA(60)a, rovA::Tn10(60)b AmpR, CmlRThis study
 pACYC184Cloning vector, p15a, CmR, TetR Chang and Cohen (1978)
 pAK001R6K derivative, inv+, ApRThis study
 pAK002pAK001, inv::TnphoA(60)a, ApR, KnRThis study
 pAK010pACYC184, rovA+, CmRThis study
 pAK011pACYC184, rovA+, CmRThis study
 pAK012pACYC184, rovA+, CmRThis study
 pGN1pPD264Δ, −426 bp inv upstream region, CmR, KnRThis study
 pGN2pPD264Δ, −328 bp inv upstream region, CmR, KnRThis study
 pGN3pPD264Δ, −207 bp inv upstream region, CmR, KnRThis study
 pGN4pPD264Δ, −50 bp inv upstream region, CmR, KnRThis study
 pGN5pPD264Δ, −31 bp inv upstream region, CmR, KnRThis study
 pGN6pPD264Δ, +23 bp inv upstream region, CmR, KnRThis study
 pGN7R6K derivative, rovA+, ApRThis study
 pGN8pPD264Δ, −90 bp inv upstream region, CmR, KnRThis study
 pGN10pQE60, rovA+, ApRThis study
 pHSG575Cloning vector, plac, CmR Takeshita et al. (1987)
 pJL29Promoter probe vector, lacZ+, ApRJ. Lucht
 pPD264pACYC184, inv::TnphoA(60)a, KnR, CmRThis study
 pPD280pGN7, rovA::Tn10(60)b, CmR, ApRThis study
 pPD281pGN7, rovA–lacZ(129)c, ApRThis study
 pPD285pPD280, sacB+, ApR, CmRThis study
 pPD293pHSG575, plac::rovA+, CmRThis study
 pQE60Expression vector for the synthesis of 6-His-tagged proteinsQiagen
 pRI203pBR325, inv+, ApR Isberg et al. (1987)

HEp-2 cells were cultured in RPMI-1640 media (Gibco BRL) supplemented with 5% newborn calf serum (Gibco BRL) and 2 mM glutamine at 37°C in the presence of 5% CO2.

DNA manipulations and sequence analysis

All DNA manipulations, restriction digestions, ligations and transformations were performed using standard genetic and molecular techniques (Sambrook et al., 1989;Miller, 1992). Plasmid DNA was purified using a Qiagen kit. Restriction and DNA-modifying enzymes were obtained from Boehringer Mannheim or New England Biolabs and used according to the manufacturers' instructions. The oligonucleotides used for amplification by PCR, sequencing and primer extension were purchased from Metabion. PCR reactions were routinely performed in a 100 µl mix for 25 cycles using Taq polymerase (Appligene) according to the manufacturer's instructions in a DNA thermal cycler Primus (MWG-Biotech). PCR products were purified with the QIAquick kit (Qiagen) before and after digestion of the amplification product. Radiolabelled nucleotides [α-32S]-dATP (6000 Ci mol−1) were purchased from DuPont (NEN Research Products). Double-stranded DNA sequencing of the inv upstream region was carried out by the dideoxy-chain termination procedure, using a Thermo Sequenase cycle sequencing kit according to the manufacturer's instructions (Amersham Life Science). Other sequencing reactions were performed by GATC.

Plasmids used in this study are listed in Table 1. Plasmid pAK001 was constructed by inserting a SalI–EcoRI fragment containing the inv gene of plasmid pRI203 (Isberg et al., 1987) into a mobilizable R6K-based suicide vector (Miller and Mekalanos, 1988). Plasmid pAK002 is a derivative of pAK001 carrying a TnphoA insertion in codon 60 of the inv gene generating an in frame inv–phoA fusion (for construction, see below). Plasmids pAK011–pAK013 are pACYC184 derivatives carrying chromosomal Sau3A fragments of Y. pseudotuberculosis harbouring the rovA gene. pPD264 was constructed by inserting an NruI–EcoRV fragment carrying the inv::TnphoA(60) fusion into the NruI–EcoRV site of pACYC184. pGN1–pGN6 and pGN8 were constructed by inserting PCR fragments harbouring different portions of the inv promoter region into pPD264 cut with BamHI and SwaI (Fig. 5A). The products were generated using the universal downstream primer 5′-ATTCCTTATCAAGAGAAACTCACT-3′ and the upstream primers 5′-GCGCGGATCCATTTTGGGTGAACACAG-3′ (pGN1), 5′-GCGCGGATCCACTGTTTAGTTCACGATGG-3′ (pGN2), 5′-GCGCGGATCCGACAACATCAAATGAAGG-3′ (pGN3), 5′-GCGCGGATCCTTTTAATCTGCCCAGTCTGG-3′ (pGN4), 5′-GCGCGGATCCGGTTTTTTAAAAAAGTGCTATCC-3′ (pGN5), 5′-GCGCGGATCCGTTTCACTTTCAACTATTGC-3′ (pGN6) and 5′-GCGCGGATCCGCTGTATGGTATAGACTG-3′ (pGN8) and cut with BamHI (created in upstream primer) and SwaI (located in the inv upstream region). pGN7 was constructed by inserting a SalI–EcoRV fragment containing the rovA gene of plasmid pAK012 into a R6K derivative constructed by Miller and Mekalanos (1988) cut with SalI and EcoRV. To generate and overexpress RovA-6His, a PCR fragment carrying the entire rovA gene was inserted into the BamHI site of pQE60, using primers 5′-GCGCGGATCCGAATCGACATTAGGATCTGATC-3′ and 5′-GCGCGGATCCCTTAGTTTGTAATTGAATATTATT-3′, both creating BamHI sites. The resulting plasmid pGN10 encoded a C-terminal His-tagged RovA protein, which was synthesized in E. coli strain XL1blue (see below). pPD281 was constructed by ligating the SmaI–DraI fragment of pJL29 carrying the lacZ gene into the StuI site in rovA on plasmid pGN7. The junction of the rovA and the lacZ genes and the regulatory region of rovA were confirmed by DNA sequencing. Plasmid pPD293 was constructed by inserting an EcoRI–SalI PCR fragment containing the rovA gene downstream of the lac promoter of vector pHSG575 (Takeshita et al., 1987). The PCR product was generated by the upstream primer GCGGCGGAATTCGACTGGACTCTGGGAGTTATGC and the downstream primer CGCGCGGTCGACGGCTTGTTTTGTTCTGCTGTG creating an EcoRI and SalI site.

Construction of the Y. pseudotuberculosis gene bank, isolation of sequences complementing inv–phoA fusion expression in E. coli

For the construction of a Y. pseudotuberculosis gene bank, chromosomal DNA of Y. pseudotuberculosis pIB1 was prepared as described previously (Silhavy et al., 1984). Fragments (2–8 kb) generated by a partial Sau3A digest were purified and ligated in pACYC184 predigested with BamHI. The plasmid library was tested by restriction digests and introduced into CC118λpir/pAK002 carrying an inv–phoA fusion. Twenty clones showing an elevated phoA expression on plates supplemented with chloramphenicol, ampicillin and X-P were isolated. The gene bank plasmids were reisolated and retransformed into CC118λpir/pAK002 to verify their influence on inv–phoA expression. Three of the identified clones, pAK011–pAK013, also exhibited a much higher inv expression rate in Y. pseudotuberculosis and were used for further analysis.

Construction of the chromosomal inv–phoA and the rovA–lacZ translational fusions

To isolate invasin fusions to the phoA indicator gene, a Tn5 IS50L::phoA (TnphoA) derivative was inserted into the mobilizable inv+ plasmid pAK001, which was not able to replicate in Y. pseudotuberculosis, as described previously (Manoil and Beckwith, 1986;de Lorenzo et al., 1990). Active phoA+, KanR colonies were selected, and insertions in the inv gene were physically mapped by restriction analysis and sequenced. One of the plasmids (pAK002) carrying the inv::TnphoA(60) insertion, generating an in frame inv–phoA fusion after codon 59 of invasin, was propagated in E. coli S17-1 (tra+) in the presence of the pi protein and transferred by the RP4 mob functions of pAK002 into Y. pseudotuberculosis YPIII. Y. pseudotuberculosis transconjugants were selected on Yersinia selective agar (Oxoid) supplemented with antibiotics, Amp and Cm, selecting for the drug resistance of the plasmid. Inasmuch as Y. pseudotuberculosis cannot replicate R6K derivatives, strains that arise are the result of plasmid integration into the Yersinia chromosome by homology provided by the inv locus. The recombination event yielded a merodiploid strain YP1, which includes a wild-type copy of inv and the inv–phoA(60) fusion. A spontaneous second recombination process resulted in the excision of the integrated plasmid including the inv+ gene, leaving the inv::phoA fusion in the chromosome. The resulting strains YP1 (inv+, inv–phoA) and YP2 (inv, inv–phoA) were subsequently analysed by Southern blot hybridization and DNA sequencing to determine whether the chromosome of these recombinants had the appropriate genetic structure. For the construction of the chromosomal rovA–lacZ translational fusion strains, YP4 and YP6, plasmid pPD281 was mated from E. coli S17-1λpir (tra+) into Y. pseudotuberculosis YPIII and YP3 as described above. The isolated transconjugants on Yersinia selective agar (Oxoid) supplemented with Amp were screened for stable chromosomal insertion into the rovA locus by PCR.

Construction of the rovA::Tn10(Cm) insertion

For the construction of a rovA::Tn10 insertion mutant, a Tn10(CmR) derivative was inserted into the mobilizable inv+ plasmid pGN7 as described previously (de Lorenzo et al., 1990). Nine independent pools of mutagenized plasmids were generated. Five CmR colonies were selected from each pool and tested for RovA function by investigating their inability to induce a plasmid-encoded inv–phoA fusion in E. coli. Functionless insertions in the rovA gene were physically mapped and sequenced. In one of the mutant plasmids pPD280, carrying the Tn10(CmR) insertion in codon 60 [rovA::Tn10(60)], a SalI fragment harbouring the sacB gene of pAY01 (Yang and Isberg, 1993) was inserted into the unique SalI site. The resulting plasmid pPD285 was subsequently mated from E. coli S17-1λpir (tra+) into Y. pseudotuberculosis YPIII as described above. Y. pseudotuberculosis transconjugants were selected on Yersinia selective agar (Oxoid) supplemented with Cm. The recombination of the plasmid into the Yersinia chromosome yielded a merodiploid strain, which includes a wild-type copy of rovA and the rovA::Tn10(60) insertion. Subsequently, the resulting strain was plated on 10% sucrose and chloramphenicol, and fast-growing, large colonies were selected. Because sucrose induces the expression of the sacB gene on the integrated plasmid pPD285 and leads to the production of a toxic substance that prevents growth (Gay et al., 1985), a spontaneous second recombination process that results in the excision of the integrated plasmid including the rovA wild-type copy is advantageous. One hundred selected CmR strains were screened for ampicillin sensitivity to prove the loss of the integrated plasmid, and 14 independent Y. pseudotuberculosis mutants carrying the rovA::Tn10 mutation on the chromosome were maintained. One strain, YP3, harbouring the rovA::Tn10 mutation in the rovA locus, as proved by PCR and DNA sequencing, was taken for further analysis.

Nucleotide sequence accession numbers

The GenBank accession number for the Y. pseudotuberculosis flhE–inv region is AF330138. The GenBank accession number for the complete sequence of rovA from Y. pseudotuberculosis is AF330139.

Primer extension analysis

Total RNA was prepared by hot acid–phenol extraction from cultures of Y. pseudotuberculosis YPIII or YPIII harbouring pPD264 (inv–phoA) grown to exponential or stationary phase at 25°C and 37°C respectively. RNA concentration and quality were determined by measurement at A260 and A280. Primer extension analysis reaction was performed by modifying the procedure reported by Dersch et al. (1993). For each primer extension reaction, the synthetic primer 5′-ATTCCTTATCAAGAGAAACACAC-3′ for the inv gene was hybridized with 50 µg of RNA and extended with AMV reverse transcriptase in the presence of radiolabelled [α-35S]-dATP. Sequence ladders were generated with the same primer using plasmid pPD264 (inv–phoA) by double-stranded sequencing as described previously (Sambrook et al., 1989). The size of the reaction product was determined on a denaturing 4% DNA sequencing gel.

Expression and purification of the RovA protein

XL1blue/pGN10 (5 l) was grown at 37°C in LB broth to an A600 of 0.6. IPTG was added to a final concentration of 2 mM to induce the expression of RovA-6His. The cells were grown for an additional 2 h before being harvested. Frozen cell pellets were resuspended in 50 ml of 50 mM NaH2PO4, pH 8.0, and 300 mM NaCl, 10 mM imidazole, 50 mg lysozyme was added and incubated on ice for 1 h. Subsequently, the bacteria were lysed by sonication (Branson Instruments, 50% pulse for 5 min). The soluble RovA-6His extract was separated from insoluble cell material by centrifugation at 25 000 g and purified by affinity chromatography on Ni-TA agarose (Qiagen). The column was washed with three column volumes of 50 mM NaH2PO4, pH 8.0, 300 mM NaCl, 40 mM imidazole and eluted with 50 mM NaH2PO4, pH 8.0, 300 mM NaCl containing 250 mM imidazole. The eluted RovA-6His protein was concentrated with Centricon 10 (Amicon) and dialysed against the reaction buffer for the gel retardation assay (see below).

DNA retardation assay

The binding of RovA to defined PCR fragments carrying different portions of the inv regulatory region was carried out in a 20 µl reaction mixture containing increasing amounts of purified RovA protein (0.8–1.6 µg) and 1 µg of DNA. The reaction buffer contained 10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 50 mM NaCl, 5 mM MgCl2, 5 mM dithiothreitol (DTT) and 5% glycerol. The reaction mixture were incubated for 20 min at room temperature and subsequently loaded on to a 4% polyacrylamide gel. The gel was run in 1× TAE (Sambrook et al., 1989) and stained with ethidium bromide. PCR fragments of the assays were amplified with primers used for the construction of pGN1–6 and 8. The csiD promoter fragment was amplified with primers 5′-CATTTCCTGTCTCACTTTGCC-3′ and 5′-GTCCTGGCCTGAATCGACAGCG-3′.

β-Galactosidase and alkaline phosphatase assays

β-Galactosidase and alkaline phosphatase activities were measured in permeabilized cells slightly modified as described previously (Manoil and Beckwith, 1986; Miller, 1992). For the alkaline phosphatase assays, 2 ml of the cultures was washed twice with 250 mM Tris-HCl to remove remaining traces of phosphate in the medium; for β-galactosidase assays, the bacteria were resuspended in 2 ml of 1× Z buffer (Miller, 1992). To lyse the cells, 40 µl of 0.1% SDS and 80 µl of chloroform were added to the cultures and incubated for 10 min at room temperature. The reaction was subsequently started by adding 400 µl of PNPP and ONPG (Sigma), respectively, and stopped with 1 ml of 1 M NaCO2 or K2HPO4 after the samples turned yellow. The activities were calculated as follows: β-galactosidase activity OD420 × 6.75 × inline image × Δt (min)−1 × Vol. (ml)−1; and alkaline phosphatase activity OD420 × 6.46 × inline image × Δt (min)−1 × Vol. (ml)−1. β-Galactosidase assays were performed in triplicate cultures grown under the conditions indicated. Wild-type Y. pseudotuberculosis YPIII assayed under identical conditions had a low detectable background level of β-galactosidase and alkaline phosphatase activity, which was subtracted from the values presented.

Gel electrophoresis, preparation of cell extracts and Western blotting

Cultures of Y. pseudotuberculosis were grown under various environmental conditions as described above. The optical density of the cultures was adjusted, and a 1 ml aliquot was withdrawn from each culture. The cells were collected by centrifugation, resuspended in 100 µl of sample buffer (Sambrook et al., 1989) and subsequently lysed at 95°C for 5 min. To reduce the viscosity of the whole-cell extracts, 3 µl of benzonase (Merck) was added to the samples and incubated at 37°C for 10 min. For the immunological detection of the invasin and the RovA–LacZ fusion proteins, 5 µl portions of the cell extract were loaded onto 10% SDS–polyacrylamide gels, and the proteins were separated by electrophoresis and transferred onto an Immobilon membrane (Millipore). The bound proteins were then probed with either a monoclonal antibody (3A2) directed against invasin (Dersch and Isberg, 1999) or a polyclonal LacZE. coli-specific antiserum. The antigen–antibody complexes were visualized with a second goat alkaline phosphate antibody (Sigma) using 5-bromo-4-chloro-3-indoylphosphate and nitroblue tetrazolium (Boehringer Mannheim) as substrates.

Cell adhesion and invasion assay

For preparation of the cell adhesion and uptake assay, 5 × 104 HEp-2 cells were seeded and grown overnight in individual wells of a 24-well cell culture plates (Nunc). Cell monolayers were washed three times with PBS and incubated in RPMI-1640 medium supplemented with 20 mM HEPES (pH 7.0) and 0.4% BSA before the addition of ≈ 5 × 106 bacteria. Bacteria were added to the monolayer and incubated at 20°C to test for cell binding (and prevent uptake) or at 37°C to test for invasion. To assay cell binding, 1 h after infection, the cells were washed extensively with PBS five times. The total number of adherent bacteria were determined by cell lysis using 0.1% Triton and plating on bacterial media. Bacterial uptake was assessed 90 min after infection as the percentage of bacteria that survived killing by the addition of the antibiotic gentamicin to the external medium, as described previously (Dersch and Isberg, 1999). For each strain, the relative level of bacterial adhesion/uptake was determined by calculating the number of cfu that arose relative to the total number of bacteria introduced onto monolayers, performed in triplicate.


We particularly thank Tanja Heise, Ann-Kathrin Heroven, Annette Krechel and Susanne Parr for technical assistance. We also thank Drs Birgitta Beatrix, Martin Fenner, Jacinta Lodge and Eckhard Strauch for helpful discussions and critical reading of the manuscript. We thank Dr R. Hengge-Aronis, in whose laboratory this work was carried out, for her generous support, interest and advice. This work was supported by grant DE 616/2-1 from the Deutsche Forschungsgemeinschaft to P. Dersch, and the Gottfried-Wilhelm Leibniz award to Professor Dr R. Hengge-Aronis.