A chromosomally encoded regulator is required for expression of the Yersinia enterocolitica inv gene and for virulence



The primary invasion factor of Yersinia enterocolitica, invasin, is encoded by inv. inv expression is regulated in response to pH, growth phase and temperature. In vitro, inv is maximally expressed at 26°C, pH 8.0, or 37°C, pH 5.5, in early stationary phase. At 37°C, pH 8.0, inv is weakly expressed. To identify which gene(s) are required for inv regulation, we screened for transposon insertions that decreased expression of an inv–′phoA chromosomal reporter at 26°C. Of 30 000 mutants screened, two were identified that had negligible inv expression in all conditions tested. Both of these independent mutants had an insertion into the same gene, designated rovA (regulator of virulence). RovA has 77% amino acid identity to the Salmonella typhimurium transcriptional regulator SlyA. Complementation with the wild-type rovA allele restores wild-type inv expression as monitored by Western blot analysis, tissue culture invasion assay and alkaline phosphatase assay. There is also a significant decrease in invasin levels in bacteria recovered from mice infected with the rovA mutant; therefore, RovA regulates inv expression in vivo as well as in vitro. In the mouse infection model, an inv mutant has a wild-type LD50, even though the kinetics of infection is changed. In contrast, the rovA mutant has altered kinetics, as well as a 70-fold increase in the LD50 compared with wild type. Furthermore, because the rovA mutant is attenuated in the mouse model, this suggests that RovA regulates other virulence factors in addition to inv. Analysis of other proposed virulence factors such as Ail, YadA and the Yop proteins shows no regulatory role for RovA. The more severe animal phenotype combined with the lack of impact on known virulence genes aside from inv suggests RovA regulates potentially novel virulence genes of Y. enterocolitica during infection.


Yersinia enterocolitica is a Gram-negative enteric human pathogen that causes several diseases, primarily in children and young adults. The most common manifestation of Y. enterocolitica infection is enterocolitis. Other less common manifestations include mesenteric lymphadenitis and septicaemia (Bottone, 1997). Y. enterocolitica is usually acquired by ingesting contaminated food or water. Upon ingestion, the organisms must pass through the stomach to the small intestine, where they bind to and invade the epithelium of the terminal ileum.

As with many enteric pathogens, tissue invasion is a critical first step in the pathogenesis of Y. enterocolitica infections. Factors important for the initial step of invasion include YadA (encoded on the Yersinia virulence plasmid; Wachtel and Miller, 1995), and invasin (encoded on the chromosome; Miller and Falkow, 1988). Invasin is a 92 kDa outer-membrane protein that is closely related to the Y. pseudotuberculosis invasin (Isberg and Falkow, 1985). The Yersinia pseudotuberculosis homologue was shown to bind to host cell β1 integrins (Isberg and Leong, 1990), which are expressed on the apical surface of the antigen-sampling M-cells that overlie the Peyer’s patches (Clark, 1998). The binding of invasin to these cellular proteins results in the active invasion of the host cells in culture. The available evidence also suggests that invasin-mediated binding to these integrins on the M-cells promotes invasion of the intestinal mucosa by the bacteria. Once the organisms invade the M-cells, they translocate through to the Peyer’s patches and efficiently colonize these tissues. After this initial translocation and colonization, the bacteria spread to the mesenteric lymph nodes, via an unknown mechanism, where they replicate and eventually disseminate to other tissues including the spleen, liver and lungs (Pepe and Miller, 1993).

Invasin is important early in the infection process, as was demonstrated by the phenotype of an inv mutant in the mouse infection model; the inv mutant does not efficiently colonize the Peyer’s patches early in infection (Pepe and Miller, 1993). At later stages in the infection, however, the Peyer’s patches are colonized, and the organisms eventually spread to and colonize the deeper tissues. This delayed colonization appears to require the secondary invasion factor YadA (Pepe et al., 1995). The inv mutant is not altered in the oral or intraperitoneal LD50 (Pepe and Miller, 1993). These data suggested that invasin is involved in the initial translocation through the M-cells to the Peyer’s patches, but is not required for survival in or dissemination to deeper host tissues.

Invasin is clearly an important virulence determinant of Y. enterocolitica, thus a better understanding of the environmental signals triggering inv expression and of the factors regulating inv expression would be of interest. Pepe et al. (1994) showed that in the laboratory inv is expressed maximally at 26°C and neutral pH, and is only weakly expressed at 37°C and neutral pH. As the pH of the growth medium is lowered, inv expression is increased in cultures grown at 37°C. At a pH more similar to that of the distal small intestine, pH 6.0, inv expression in cultures grown at 37°C reached levels similar to those of cultures grown at 26°C at neutral pH. In addition to pH and temperature, growth phase was also found to affect inv expression. Maximal expression was reached at late exponential to early stationary phase in standard growth media (Pepe et al., 1994). Although these data show that inv expression is highly regulated, the molecular basis for these regulatory effects is largely unknown. In order to understand better the molecular mechanisms of inv regulation, a genetic search for potential regulators was performed. The result of this search was the identification of RovA, a transcriptional regulator with homology to the Salmonella typhimurium transcriptional regulator SlyA.


Mutagenesis and identification of inv regulatory mutants

A merodiploid inv–′phoA reporter strain JB41v (Badger and Miller, 1998) was mutagenized with transposon mTn5Km2 (Hensel et al., 1995) to identify the factor(s) responsible for inv regulation. Approximately 30 000 transposon insertion mutants from four independent matings were screened at 26°C on Luria–Bertani (LB) plates containing the chromogenic indicator XP. JB41v colonies are blue on these plates when grown at 26°C because of inv–′phoA expression, and insertions that alter expression from the inv promoter can be visualized because of a change in colony colour. Five hundred mutants that had altered AP levels by this qualitative screen were purified and analysed in a quantitative alkaline phosphatase assay; 41 mutants that had < 50%, or > 100% of wild-type AP levels were analysed further.

Previous analysis showed that many of the mutations that have subtle effects on inv expression also affect motility or phospholipase (PLA) expression (Badger and Miller, 1998; data not shown). Analysis of these mutations suggested they indirectly affected inv regulation. In order to limit this search to insertion mutations that are likely to impact inv regulation directly, the mutants were checked for their motility and PLA phenotypes. The 41 mutants fell into four different classes based on the phenotypes assayed. Eight mutants were class 1, with increased AP levels, normal motility and wild-type PLA activity. Seven mutants were class 2, with increased AP levels, altered motility but wild-type PLA activity. Nine mutants were class 3, with decreased AP levels, normal motility and wild-type PLA activity (of these nine, six were found to have insertions in either the inv promoter or the phoA gene). Seventeen mutants were class 4, with decreased AP levels and altered motility. Two of these mutants also had altered PLA activity; one of these mutants, 1b.23, was complemented for PLA and motility by the addition of the flagellar master regulator operon flhDC (Young et al., 1998) in trans (data not shown). This is consistent with the observation that in Y. enterocolitica both motility and PLA require flhDC for expression (Young et al., 1999). All of these mutants have wild-type growth rates in LB broth at 26°C (data not shown).

Class 3 mutants 3c.4 and 4a.23 were chosen for further characterization because they had the largest alteration in AP levels compared with the wild type, but still had wild-type PLA and motility phenotypes. Strains 3c.4 and 4a.23 were designated YVM534 and YVM535 respectively. Southern blot analysis showed that each mutant had only one transposon insertion (data not shown). The inv–′phoA fusion of the reporter strain allowed characterization of inv expression by alkaline phosphatase assay. AP levels of YVM534 and YVM535 were reduced to 10% of the parental levels (Table 1). Furthermore, invasin levels could be monitored indirectly by tissue culture invasion assay and directly by Western blot analysis. The ability of YVM534 and YVM535 to invade Chinese hamster ovary (CHO) cells was greater than 20-fold reduced relative to JB41v (Table 1). Invasin levels in YVM534 and YVM535 were nearly undetectable by Western analysis of cultures grown under any of the conditions known to induce expression in vitro (i.e. 26°C, pH 8.0; 26°C, pH 5.5; 37°C, pH 5.5) (Fig. 1A and B; it is typical to see a large number of breakdown products in addition to full-length invasin). Taken together, these data suggest that the transposon insertions in YVM534 and YVM535 disrupt the factor(s) involved in inv regulation in vitro and that the disrupted factor(s) is trans acting because it affects expression of both inv and inv–′phoA expression.

Table 1. Invasion phenotype and expression of invphoA reporter.
 Percentage invasiona (% of wild-type level)bRelative AP activityc
StrainWithout vectorWith vectordWith rovAdWithout vectorWith vectordWith rovAd
  1. a . Percentage invasion of CHO cells was determined using the method described in Experimental procedures.
    b . Percentage of wild-type level was determined by normalizing percentage of JB41 invasion to 100%.
    c . Relative AP activity is reported as percentage of wild-type AP activity, with JB41 AP levels reported as 100%.
    d . The vector was pWKS30strep and the rovA & p lus; plasmid was pRev9.

JB418 (100)10 (125)11 (138)10099170
YVM5340.06 (0.75)0.11 (1.4)12 (150)911100
YVM5350.10 (1.25)0.15 (1.9)9 (113)1113155
Figure 1.

Comparison of Inv levels expressed by wild-type (JB41) and by the rovA mutant (YVM534). Western blot analysis of cellular proteins from JB41 and YVM534 using polyclonal anti-invasin antibodies.

A. Samples grown in different conditions known to induce inv expression, as indicated above each lane. Each sample was loaded with 0.25 OD equivalents.

B. Samples grown at 26°C neutral pH with or without the rovA complementing clone supplied in trans. The vector was pWKS30strep, and the rovA+ plasmid was pRev9.

Characterization of mutants YVM534 and YVM535

To characterize better these two insertion mutants, we cloned and sequenced the gene disrupted by the transposon in YVM534 and YVM535, as described in Experimental procedures. YVM534 and YVM535 had insertions in the same gene, which we designate rovA (regulator of virulence). The predicted rovA open reading frame (ORF) shares 77% amino acid identity with the Salmonella typhimurium transcriptional regulator SlyA (Libby, 1994). rovA is followed by a divergently transcribed gene; therefore, it is unlikely that the insertions in YVM534 and YVM535 are polar. However, a clone that contains only the rovA ORF and surrounding non-coding sequence was constructed to address this issue more carefully. This construct pRev9, and not the vector alone, restored wild-type inv expression levels as measured using AP assay, and restored invasin levels as measured indirectly, using tissue culture invasion assay, and directly, using Western blot analysis. (Table 1, Fig. 1B). These results show that rovA is critical for inv expression in vitro.

To determine whether or not rovA regulates inv expression in vivo, we monitored invasin levels in bacteria recovered from Peyer’s patches of infected mice. BALB/c mice were orally infected with 109 colony-forming units (cfu) of JB41v or YVM534. Mice were humanely sacrificed and Peyer’s patches were harvested on day 5 after infection. Bacteria recovered from the Peyer’s patches were immediately analysed by Western blot for invasin levels. JB41v isolated from Peyer’s patches expressed inv in vivo as previously shown (Pepe et al., 1994). However, invasin levels in the rovA mutant were less than 10% of the levels seen with JB41v (Fig. 2), indicating that rovA is required for normal inv expression in vivo as well as in vitro.

Figure 2.

Inv expression levels during infection of BALB/c mice. Western blot analysis of wild-type (JB41) and of the rovA mutant (YVM534) recovered from Peyer’s patches of infected mice using anti-invasin antibody. Viable counts recovered from the tissues were determined, and each lane represents a dilution of the original sample.

Effect of rovA on pathogenesis

Previous studies revealed that a lack of invasin during infection results in delayed infection of the Peyer’s patches (Pepe and Miller, 1993). Therefore, an in vivo kinetic analysis was performed to address whether the reduced invasin levels in the rovA strain would result in a similar alteration of infection kinetics. Mice were infected with the parental strain (JB41v), the inv mutant (JP273v) or the rovA mutant (YVM534); both a high dose (1 × 109) and a low dose (5 × 107) were tested. At various times after infection, mice were dissected and the number of bacteria present in the Peyer’s patches, mesenteric lymph nodes or spleen was determined by plating on LB agar plates containing nalidixic acid to select for Y. enterocolitica. Bacterial load was recorded as cfu g− 1 of tissue. The results of the infection with 109 cfu showed that the rovA mutant is capable of establishing an infection and replicating in the Peyer’s patches (Fig. 3A). Although fewer bacteria were recovered from the Peyer’s patches of mice infected with the rovA mutant than from Peyer’s patches of mice infected with the parent strain, colonization of these tissues at 24 h by rovA was more consistent than colonization by the inv mutant (Fig. 3A). Overall, the Peyer’s patches of the rovA-infected mice were significantly less inflamed than those from mice infected with either the wild type or the inv mutant, even when equivalent numbers of bacteria were present (data not shown). More striking was the difference in the dissemination of the bacteria. The rovA strain was attenuated in its ability to reach the deeper tissues, although it occasionally reached the spleen. Additionally, the route of spread to the deeper tissues appeared to be altered because colonization of the spleen did not seem to require prior colonization of the mesenteric lymph nodes (Fig. 3B and C). Instead, in some mice it looked as if the rovA strain spread directly from the Peyer’s patches to the spleen. Furthermore, in the few instances in which the organisms reached the spleen, there was significant macroscopic abscess formation (data not shown) and a very high bacterial load was recovered. The reduced ability to colonize the mesenteric lymph nodes and the spleen was not reproduced in an intraperitoneal infection. When mice were inoculated by this route, there was no difference in the ability of the rovA strain and of the parent strain to survive or replicate in the deeper tissues (data not shown). This shows that the significant decrease in bacteria recovered from the MLN and the spleen after oral infection is not due simply to an inability to grow in these tissues and suggests a defect in dissemination or initial colonization of the deeper tissues by the rovA mutant. The difference in infectious kinetics of the rovA mutant strain compared with the wild type or the inv mutant suggests that the normal pathogenesis of Y. enterocolitica infection is more significantly altered by the loss of rovA than by the loss of inv alone.

Figure 3.

Kinetics of infection for wild-type (JB41), the inv mutant (JP273) and the rovA mutant (YVM534). Viable counts recovered from tissues of infected mice were recorded as cfu g− 1. Each bar represents data from one animal.

A. Number of cfu g− 1 recovered from Peyer’s patches (PP) of mice infected with 109 cfu 24 h after infection.

B. Number of cfu g− 1 recovered from PP, mesenteric lymph nodes (MLN) and spleen at day 6 after infection with 109 cfu.

C. Number of cfu g− 1 recovered from PP, MLN and spleen at day 6 after infection with 5 × 107 cfu.

Previous studies indicated that the oral LD50 of the inv null strain is the same as that of the wild-type (Pepe and Miller, 1993). To assess further the impact of the rovA mutation on pathogenesis, we did an oral LD50 analysis with the rovA mutant. Five groups of six mice were orally infected with different doses of wild-type (JB41v) or of the rovA mutant YVM534. The LD50 for mice infected with JB41v and YVM534 was 1 × 106 and 7 × 107 respectively. Thus, the rovA mutant is 70-fold attenuated compared with wild-type and thus the inv mutant. These data strongly suggest that in addition to inv RovA regulates other genes important for Y. enterocolitica virulence.

Impact of rovA on the regulation of known virulence genes

Because of the evidence that RovA regulates virulence factors other than inv, the effect of a mutation in rovA on expression of other known virulence genes was assessed. The most thoroughly characterized Y. enterocolitica virulence factors are the Yop proteins that are encoded on the large Yersinia virulence plasmid pYV. yop expression is induced in vitro at 37°C in calcium-depleted media, whereas inv expression is low under these growth conditions (Cornelis et al., 1998). These observations make it unlikely that RovA is involved in yop expression, but do not eliminate the possibility. To determine whether the loss of rovA alters yop expression, secreted proteins from cultures grown at 37°C with limiting calcium were analysed using SDS–PAGE (Fig. 4A). The loss of rovA does not affect yop expression as the protein profiles of the rovA mutant were identical to the protein profiles of the wild type. Another virulence gene encoded by pYV is yadA; its expression is also induced at 37°C. To address the possibility that RovA regulates yadA expression, proteins from bacteria grown at 37°C were analysed by SDS–PAGE and, as with the Yop proteins, the loss of RovA had no effect on yadA expression (Fig. 4B). Finally, expression of the chromosomally encoded ail gene, which is also induced at 37°C, was not affected by the loss of RovA (data not shown). Although RovA appears to regulate factors in addition to inv that are important for Y. enterocolitica pathogenesis, these factors do not include several of the known virulence genes. This suggests that RovA may be regulating previously unidentified genes or known genes whose relevance during the infectious process has not been previously described.

Figure 4.

Analysis of Yop protein secretion and yadA expression in wild-type (JB41) and in the rovA mutant strain (YVM534).

A. YadA levels were compared between the wild-type and the rovA mutant strains. Whole cell pellets from overnight cultures grown at 26°C or 37°C were separated by SDS–PAGE. YadA is marked with the labelled arrow.

B. Yop proteins were harvested and analysed as described in the Experimental procedures. The first two lanes are wild-type (JB41) and the rovA mutant (YVM534) strains that carry the virulence plasmid. The last two lanes are the plasmid-cured counterparts. Molecular weight standards are indicated on the far left.


In order to cause infection, Yersinia enterocolitica must effectively cross the epithelial barrier of the terminal ileum. The organisms do this by actively invading through the M-cells that overlie the Peyer’s patches. The primary invasion factor involved in this initial invasion of the host epithelial barrier is invasin (Pepe and Miller, 1993). Invasin is known to be important in the early stages of the infection process because an inv mutant is delayed in the initial colonization of the host Peyer’s patches (Pepe and Miller, 1993).

The invasin gene inv is expressed in the Peyer’s patches of infected animals as well as in vitro under a variety of different growth conditions (Pepe et al., 1994). Although inv expression is highly regulated, the mechanism of this regulation was not well understood. Through transposon mutagenesis, one factor required for inv expression in vitro was identified and designated rovA for regulator of virulence. The screen in which rovA was identified was designed to identify factors that are involved in the regulation of inv expression in Y. enterocolitica grown in LB at neutral pH at 26°C. Wild-type Y. enterocolitica grown in these conditions expresses high levels of invasin. Strains with a transposon insertion in the rovA gene were identified by the significant decrease in inv expression compared with the wild type when grown under these conditions. Although inv is not expressed at high levels in LB (pH 8.0) at 37°C, expression is induced in LB (pH 5.5) at 37°C. RovA is required for this induction as well as for the induction at 26°C (neutral pH). The requirement of rovA for inv expression under all in vitro conditions known to induce inv expression indicates that rovA is critical for inv expression in vitro. Importantly, analysis of Y. enterocolitica recovered from Peyer’s patches of infected mice revealed that rovA is also required for inv expression in vivo. This has rarely been directly demonstrated for virulence gene regulators.

Sequence analysis revealed that RovA has 77% amino acid identity to the Salmonella transcriptional regulator SlyA (Libby, 1994). SlyA is a member of the MarR family of transcriptional regulators, which was first identified by its ability to activate expression of a cryptic haemolysin in Escherichia coli. The MarR family encompasses a number of bacterial proteins that have a range of regulatory functions. The more distantly related family members include MarR, ErmR and PecS. MarR and ErmR are negative regulators of antibiotic resistance systems in E. coli (Miller, 1996). PecS negatively regulates pectinase and cellulase production in the plant pathogen Erwinia chrysanthemi (Reverchon, 1994). RovA is more closely related to SlyA than to any of the other members of this family.

Each of these RovA homologues is believed to regulate multiple genes. In particular, SlyA and PecS are thought to regulate multiple virulence genes. To address the possibility that rovA regulates virulence genes in addition to inv, the rovA mutant was tested in the mouse infection model. The prediction was that if RovA regulates inv expression but no other virulence genes then a rovA mutant should have a phenotype similar to that of the inv mutant in the mouse model. Kinetic analysis revealed that the rovA mutant has a phenotype distinct from either wild-type or the inv mutant. The rovA mutant colonized the Peyer’s patches more efficiently than the inv strain, but less efficiently than wild-type. Moreover, the rovA mutant does not colonize the deeper tissues, including the mesenteric lymph nodes and spleen, as well as either wild-type or the inv mutant. This defect in colonization of the deeper tissues by the rovA strain suggested that factors in addition to inv are regulated by RovA. To test further the difference in pathogenesis among rovA, inv and wild-type strains, we carried out LD50 analysis. The oral LD50 of the rovA strain is 70-fold higher than that of the inv strain or of the wild-type strain. Together, these data strongly support the hypothesis that virulence genes in addition to inv are regulated by RovA. Analysis of several known Y. enterocolitica virulence factors that are normally induced at 37°C, including YadA, Ail, and the Yop proteins, revealed that RovA did not affect their production. Furthermore, RovA did not impact upon motility or phospholipase activity, both of which are induced in conditions that induce inv expression.

Interestingly, the regulation of inv by RovA may be yet another example of an endogenous regulator regulating the expression of an acquired virulence gene. The position of inv interrupting a flagellar locus and the relatively low G + C content of this gene suggest that inv may have been acquired by horizontal gene transfer. Other examples of this phenomenon include the regulation of ctxAB (cholera toxin) by toxR of Vibrio cholerae (Faruque et al., 1998) and the regulation of some of the genes in Salmonella typhimurium pathogenicity island 1 by phoP/phoQ (Pegues et al., 1995).

In Salmonella, the slyA mutant is highly attenuated with a greater than 1000-fold increase in the oral LD50 and a greater than 10 000-fold increase in the intraperitoneal LD50 in the mouse infection model (Libby, 1994). Although the degree of attenuation for the rovA strain of Y. enterocolitica was less than that of the slyA strain of Salmonella, there may still be a similar function for the two regulators. SlyA is thought to be involved in regulating factors involved in tolerating oxidative stress as a result of the observation that the slyA mutant is hypersusceptible to reactive oxygen intermediates (Buchmeier, 1997). Not surprisingly, given this sensitivity, the slyA mutant is defective in its ability to survive inside macrophages (Libby, 1994). Preliminary data suggest the rovA mutant strain is also hypersensitive to hydrogen peroxide exposure (S. Libby, P. Revell and V.L. Miller, unpublished). Considering this, perhaps the significant difference between the in vivo phenotype of the rovA mutant and that of the slyA mutant reflects the differences between the infectious lifestyles of Y. enterocolitica and Salmonella more than the differences between the genes regulated by RovA compared with SlyA. Salmonella pathogenesis is highly dependent on the ability of the bacteria to survive within macrophages, whereas Y. enterocolitica is thought to be primarily an extracellular pathogen. Future identification and characterization of the additional genes regulated by RovA may lead to identification of virulence factors in Y. enterocolitica not previously described or perhaps to the implication of previously identified genes as important factors in Y. enterocolitica virulence. This will clarify the role that, in addition to regulating inv expression, RovA plays in the pathogenesis of Y. enterocolitica infections.

Experimental procedures

Bacterial strains and growth conditions

YVM534 and YVM535 are derivatives of the previously described JB41v [ΔyenR (RM+), inv′–′phoA, inv+] (Badger and Miller, 1998). JP273v was described previously by Pepe and Miller (1993). Escherichia coli S17-1λpir was used to conjugally transfer plasmids to Y. enterocolitica (Miller and Mekalanos, 1988). Unless specified otherwise, all Y. enterocolitica strains were grown at 26°C and all E. coli strains were grown at 37°C. All strains were grown in LB broth or on LB agar plates. Antibiotics were used at the following concentrations: ampicillin (Amp; 100 μg ml− 1), chloramphenicol (Cm; 12.5 μg ml− 1), kanamycin (Kan; 100 μg ml− 1), nalidixic acid (Nal; 20 μg ml− 1), spectinomycin (Spec; 50 μg ml− 1) and streptomycin (Strep; 50 μg ml− 1).


Y. enterocolitica JB41v was mutagenized by delivering mTn5Km2 on the previously described suicide plasmid pUT (Hensel et al., 1995). The delivery plasmid was transferred from E. coli S17-1λpir to Y. enterocolitica JB41v by conjugation as previously described (Darwin and Miller, 1999). Dilutions of the matings were plated on LB agar containing Nal (to select against the donor E. coli strain), Cm (to select for the recipient Y. enterocolitica strain), Kan (to select for transposon insertions) and the alkaline phosphatase substrate 5-bromo-4-chloro-3-indoyl-phosphate (XP; 40 μg ml− 1; to screen for alterations in inv promoter activity). Colonies that were either darker blue or paler blue than JB41v on these plates were purified for further characterization.

Tissue culture invasion assay and alkaline phosphatase assay

Invasion assays and alkaline phosphatase (AP) assays were performed using bacterial cultures grown in 2 ml of LB with appropriate antibiotics in 13 mm × 100 mm Kimax tubes on a roller for 16–18 h at 26°C. Invasion assays were performed using CHO cells as has been previously described (Miller and Falkow, 1988). Results are reported as percentage invasion =  100 × (number of bacteria resistant to gentamicin/initial number of bacteria); all assays were carried out in duplicate and mean results are reported. Alkaline phosphatase activity was measured in permeabilized cells and is presented as enzyme units per OD600 as described by Manoil and Beckwith (1985). All assays were performed in duplicate, mean results are reported with JB41v referred to as wild type; relative AP activity refers to percentage of wild-type expression in a given assay.

Motility and phospholipase assays

Motility and phospholipase activities were qualitatively determined by spotting 2 μl of a culture grown for 16–18 h at 26°C in LB broth onto appropriate plates and recording the phenotype 16–18 h later. Phospholipase activity was detected using indicator plates as described previously (Schmiel et al., 1998). Motility plates were LB with 0.3% agar or 1% tryptone with 0.3% agar (Young et al., 1998).

Western blot analysis

Cultures were grown in LB for 16–18 h to similar densities, then 0.25 OD equivalents of whole cell lysates were separated in 10% SDS–PAGE. The separated proteins were transferred to a nitrocellulose membrane for immunoblot analysis (Ausubel et al., 1992). Immunoblot analysis was performed with anti-invasin polyclonal antibodies at a 1:20 000 dilution. Binding of the primary antibody was detected by incubation with goat anti-rabbit IgG conjugated to horseradish peroxidase (Sigma) and was developed with a chemiluminescent substrate (Amersham).

Characterization of transposon insertion sites and DNA sequencing

Chromosomal DNA from the mutant strains was digested with KpnI, which cuts at one end of the transposon and leaves the kanR cassette intact. Southern blot analysis was performed using the Amersham ECL nucleic acid labelling and detection kit with the EcoRI fragment containing the kanamycin cassette from the pUT mTn5kn2 plasmid as a probe. The size of the chromosomal KpnI fragment containing the transposon was determined and all fragments of approximately that size were purified from agarose gels using a Gene Clean DNA purification kit (Bio101). This DNA was ligated into KpnI-digested pHG329 (Stewart et al., 1986) and the ligation mixture was used to transform E. coli DH5α to Amp, Kan resistance; these constructs were designated pRev3 and pRev4. The Yersinia DNA immediately flanking the transposon was sequenced using the P7 primer that anneals to the O end of the transposon (Hensel et al., 1995). DNA sequencing was carried out using an Applied Biosystems DNA sequencing system and the BigDye terminator cycle sequencing kit (applied Biosystems) according to the manufacturer’s instructions. Sequences were analysed using the blastx program available from the NCBI data base at www.ncbi.nlm.nih.gov.

Cloning of rovA+ gene

A 1.4 kb fragment of chromosomal DNA that flanked the transposon, derived from pRev3, was used to probe the Y. enterocolitica chromosomal cosmid library essentially as described by Young et al. (1998). This probe was generated using PCR with P7 (Hensel et al., 1995) and the M13 universal forward primers with pRev3 as the template. A cosmid containing the interrupted gene was identified, and a 3.2 kb KpnI fragment was subcloned into the low-copy pWKS30strep vector [pWKS30 as has been described previously (Wang and Kushner, 1991) with a Strep/Spec cassette cloned into the HindIII site of the polylinker]. This plasmid was designated pRev5. The nucleotide sequence of both strands of this locus was determined essentially as described above (GenBank accession number AF171097). Using primers designed from sequence data, rovA and its (233 bp) upstream and (157 bp) downstream non-coding sequence were amplified via PCR using recombinant Pfu (Stratagene). This rovA+ fragment was cloned into pWKS30strep and designated pRev9. The nucleotide sequence of the insert was determined to ensure no mutations were introduced during the PCR.

In  vivo expression

All animal experiments were approved by the Washington University animal studies committee, the University’s animal welfare assurance number is A-3381-01. Our protocol approval number is 1990277. To monitor inv expression in vivo, we infected virus-free 6- to 7-week-old female BALB/c mice with cultures grown for 16–18 h in LB at 26°C that had been washed and resuspended in PBS. The mice were orally infected with 2 × 109 bacteria in 0.1 ml via a 21-gauge feeding tube. Mice were killed at 5 days after infection, and Peyer’s patches from infected mice were harvested, suspended in 3 ml of PBS and homogenized. Homogenates were centrifuged at 1000 r.p.m. for 1 min and were filtered through a 5 (μm filter to remove insoluble eukaryotic tissue. Aliquots were plated to quantify viable cfu. Samples with equivalent cfus were separated using SDS–PAGE and analysed using Western blot with anti-invasin antibody as described above. Different dilutions of each sample were loaded in order to qualitatively compare expression levels. This analysis was carried out in duplicate.

LD50 and kinetics analysis

Bacteria were grown for 16–18 h in LB at 26°C. Five groups of six mice were infected orally with successive 10-fold dilutions of the bacterial suspension (104 − 108 bacteria). The mice were monitored twice daily for 14 days. This analysis was carried out in duplicate. The LD50 values were determined according to Reed and Muench (1938). Kinetic analysis was carried out by orally infecting mice with parent strain (JB41v), inv mutant (JP273v) or rovA mutant (YVM534). Two experiments were performed at two different doses of either a high dose (109 cfu) or a low dose (5 × 107 cfu). At various times after infection (1, 3 or 6 days), mice from each strain were killed and dissected. Bacterial load recovered from the infected organs was determined by plating dilutions of the macerated tissues on LB plates containing Nal, to select for Yersinia, and reported as colony forming units (cfu) g− 1 of tissue.

Analysis of secreted Yop proteins

Strains were grown in 2 ml of brain–heart infusion medium (BHI; Difco) at 26°C on a roller drum for ≈16 h. These were used to inoculate 25 ml of BHI supplemented with 5 mM EGTA and 20 mM MgCl2 in a 50 ml flask so that the optical density at 600 nm was 0.1. The cultures were incubated at 26°C with rapid aeration for 2 h and then at 37°C for 4 h to induce expression and secretion of the Yops. Three OD equivalents were harvested, and bacterial cells were removed by centrifugation. Supernatants were passed through a 0.2 μm low protein binding filter (Gelman Sciences). Proteins were precipitated by adding trichloroacetic acid to a final concentration of 10% (v/v) and incubating on ice for 1 h. The proteins were collected by centrifugation in a microfuge at 13 000 r.p.m. for 30 min at 4°C. The pellets were washed with 1 ml of ice-cold acetone. Yop proteins were resuspended in 50 μl of sample buffer with β-mercaptoethanol. Twenty-five microlitres of each sample were separated by SDS–PAGE (10% polyacrylamide) and visualized by staining with Coomassie brilliant blue.


We are thankful to Andrew Darwin and Steve Libby for helpful discussions, and thank Andrew Darwin, Amy Gort and William Goldman for critical review of the manuscript. This study was supported by National Institutes of Health Grant AI27342 awarded to V.L.M. P.R. is supported by a Lucille P. Markey predoctoral fellowship and by the Cellular and Molecular Biology training grant 5T32 GM07067.