Background

The Gram-negative bacterium Y. pseudotuberculosis is a human enteropathogen which is able to cross the intestinal mucosa through the M cells in Peyer's patches and thus infect the underlying tissues (causing ileitis and mesenteric lymphadenitis). However, in elderly or debilitated individuals (those suffering from malignancies, immunodeficiencies, chronic liver diseases or diabetes mellitus, for example), the organism frequently gains access to the bloodstream and can cause an often fatal septicemia [1, 2]. Known Y. pseudotuberculosis virulence genes are transcriptionally regulated by temperature – most probably in order to adapt to the bacterium's life cycle outside and inside the host. Regulation by the omnipresent thermal stimulus can be modulated (via a wide range of mechanisms) by signals such as pH, other ion concentrations and nutrient availability (reviewed in [3]). This allows bacterial pathogens to (i) adapt their gene transcription profiles in response to environmental cues sensed during the course of infection and (ii) express the most appropriate virulence factors at the expense of useless (or even detrimental) ones.

To date, the transcriptional gene regulation occurring when Y. pseudotuberculosis enters the human bloodstream has only been inferred indirectly from in vivo results in rodent models of infection [4, 5] and in vitro gene transcription studies. The in vitro regulation of certain Yersinia virulence loci has mainly been analyzed with respect to single growth parameter changes mimicking the environmental signals known (or assumed) to be detected by bacteria in blood, such as iron scarcity, oxygen tension and pH [6, 7]. In the present work, we have adopted an intermediate approach by comparing the overall gene expression profiles of Y. pseudotuberculosis grown in human plasma and in Luria-Bertani broth. We then compared the observed variations with those recently published for Y. pestis [8], an almost genetically identical pathogen which, however, causes plague – one of the most severe systemic infections in humans and other mammals.

Results and discussion

The genome of Y. pseudotuberculosis strain IP32953 has been recently deciphered: it contains 3,951 coding sequences (CDSs), of which 99 are borne by the virulence-associated plasmid pYV and 43 are carried by a 27-kb cryptic plasmid. Only around 49% of CDSs encode a product with a putative or proven function [9]. To gain insight into the transcriptional regulation of virulence and metabolism genes that takes place when Y. pseudotuberculosis enters and multiplies in the bloodstream, we compared the transcriptome of IP32953 grown in human plasma to the one of the same strain grown in Luria-Bertani (LB). To this end, we prepared macroarrays composed of 3,674 PCR fragments of ≈ 400-base pairs (bp), covering 96% of IP32953's CDSs and used them as described elsewhere [8] and in the Methods section. Briefly, in three independent cultures, total RNA was extracted from IP32953 cells grown in LB broth or human plasma, in the exponential or stationary phase and at 28°C or 37°C. Macroarray probing was performed three times with independently retrotranscribed and 33P-radiolabeled RNA samples from each of the eight growth combinations. After macroarray imaging, hybridization intensity data were log-transformed and normalized using a simple median normalization method. Relative data have been deposited in the Genoscript database http://genodb.pasteur.fr/cgi-bin/WebObjects/GenoScript in accordance with standards of the Microarray Gene Expression Data Society (MGED). An analysis of variance (ANOVA) was carried out independently for each gene, with the three biological factors of variation (medium, temperature and growth phase) as fixed effects. This statistical approach allowed us to evaluate the transcriptional variations induced by each factor for the dataset as a whole. Thus, three ratios (corresponding to each parameter) and associated p-values were calculated for each gene. Inter-condition transcriptional differences were considered to be statistically significant if the p-value was below 0.05. Representative macroarray hybridization results were confirmed by qRT-PCR on stored RNA samples, using the constitutively expressed YPTB0775 gene (spot ID YPO3356 and coding for the outer membrane lipoprotein NplD) as a reference (Additional file 1). Since the physiological status of the bacterium during host infection is unknown, we focused our analysis on genes regulated by the temperature and/or the medium in both the exponential and stationary phases. All Y. pseudotuberculosis transcriptional variations discussed herein were compared with those of their respective Y. pestis orthologs and are summarized in Table 1. Y. pseudotuberculosis IP32953 genes regulated at the transcriptional level by growth temperature and/or medium are listed in Tables 2 and 3.

Table 1 Y. pseudotuberculosis transcriptional variations discussed in this article compared with those recently published for Y. pestis [8]
Full size table
Table 2 Y. pseudotuberculosis IP32953 chromosomal genes (sorted by COG class [28]) that are transcriptionally regulated by growth medium and/or temperature.
Full size table
Table 3 Y. pseudotuberculosis IP32953 pYV plasmid-harbored genes that are transcriptionally regulated by growth medium and/or temperature.
Full size table

Free iron limitation is a well-known stimulus encountered by bacteria in plasma [10, 11]. As expected, IP32953 genes required for iron storage (such as the ferritin-encoding gene ftnA [12] (Fig. 1)) were found to be downregulated in plasma. Transcriptional upregulation of most iron uptake systems (along with accessory protein-encoding genes tonB, exbB and exbD) (Fig. 1) is also consistent with this condition and is in agreement with the recent findings in Y. pestis [8]. As iron is used as a cofactor by numerous enzymes (mostly when complexed with sulfur), the metal is essential for a broad range of metabolic processes. Besides activation of iron homeostasis systems, lack of iron is also expected to be associated with a dramatic decrease in the transcription of genes encoding such enzymes, with the underlying goal of lowering iron consumption. This situation is exemplified by the katA gene that encodes catalase (a ferric enzyme involved in oxidative stress defense), whose transcription is decreased in both Y. pestis and Y. pseudotuberculosis during growth in plasma (Fig. 1). However, the increase in transcription of the bio locus (required for biotin synthesis [13]) and observed in both species) suggests that differential genetic control of a subset of iron-dependent enzymes may favor supply of this metal to the pathways that are most important for bacterial survival (and thus presumably at the expense of other, less critical ones). Furthermore, the impact of transcriptional downregulation on reorientation of metabolic fluxes may be minimized by the concomitant activation of genes coding for isoenzymes that are better suited to this situation.

Figure 1
figure 1

Medium- and temperature-dependent differential expression of Y. pseudotuberculosis chromosomal genes involved in virulence and/or iron uptake & storage. Significant (p < 0.05) upshifts (yellow to red scale) or downshifts (blue scale) in individual gene transcription levels when bacteria were grown in human plasma versus LB (triangles) and/or at 37°C versus 28°C (squares) are indicated by the color scale bar. Genes encoding iron uptake/storage systems, virulence factors and their regulators are symbolized by gray, white and black arrows, respectively. Nomenclature used for gene designation correspond to the Y. pseudotuberculosis IP32953 genome annotation. Mean fold changes in transcription and p-values are indicated in Table 2.

Full size image

One example is that of the manganese- and iron-dependent superoxide dismutase genes (i.e sodA and sodB), which are Fur-activated and -repressed, respectively (Fig. 1) in both Y. pestis and Y. pseudotuberculosis. Similarly, the class Ib ribonucleotide reductase (RNR)-encoding genes (nrdHIEF) are probably important for bacterial life in plasma, since they were found to be upregulated at the expense of those in classes III (nrdDG) and Ia (nrdAB) (Table 2) – even though all three classes are equally involved in generating the synthetic precursors for DNA. The fact that only the first class is Fur-activated [14] is consistent with this observation. Similar variations have also been recorded in Y. pestis [8]. However, whereas purine/pyrimidine metabolism has been shown to be essential for Y. pestis virulence [15], the role of this metabolic pathway in the physiopathology of Y. pseudotuberculosis has not yet been investigated. Along with class 1b RNRs, more than half of the enzymes in the tricarboxylic acid cycle (TCA) are known to be catalytically iron-dependent and/or believed to be transcriptionally activated by Fur [16]. Accordingly, and in line with transcriptome data from Y. pestis, we observed that transcription of these genes fell significantly when Y. pseudotuberculosis was grown in plasma.

In contrast to the low availability of iron in blood, glucose is readily available in this biological fluid and at a higher concentration (approx. 7 mM) than in LB broth. When Y. pseudotuberculosis was cultured in plasma, genes involved in glycolysis and the upstream, sugar-supplying, phosphoenolpyruvate-dependent systems were found to be upregulated, as depicted in Fig. 2. This finding is reminiscent of an aerobic phenomenon referred to as "glucose overflow metabolism"; this consists in channeling the carbon flow towards acetate formation instead of citrate formation, in order to prevent the excessive accumulation of NADH that would otherwise result from very high glucose consumption rates [17]. However, one main feature of glucose overflow in E. coli is acetate accumulation due to a strong transcriptional repression of the glyoxylate shunt aceBAK operon [18]. Interestingly, at least the first two of these genes are not down- but are up-regulated in Y. pseudotuberculosis (Fig. 2, Additional file 1), suggesting a need for this species to limit acetate overloads. The continuous de-repression of these genes (due to inactivation of the IclR repressor) suggests that this might also be the case in Y. pestis. These pathways are controlled by complex and finely balanced networks involving numerous pleiotropic regulators, including Fur, Crp, Fnr and ArcA [16, 19]. This unexpected upregulation may well result from the combination of both high glucose and low iron levels in plasma. Whether this occurs through the strong transcriptional repression observed with both fnr and arcA remains to be addressed in future experiments.

Figure 2
figure 2

Medium-dependent differential expression of genes coding for enzymes putatively involved in Y. pseudotuberculosis glycolysis and the tricarboxylic acid cycle (TCA cycle). Significant (p < 0.05) upshifts (yellow to red scale) or downshifts (blue scale) in individual gene transcription levels in human plasma versus LB is indicated by the color scale bar. Open boxes indicate genes whose expression levels did not vary significantly (p > 0.05). Although considered as not significant by statistical analysis of macroarray data (p = 0.053), transcriptional upregulation of aceB in human plasma was confirmed by qRT-PCR. Abbreviations: Ac-CoA: acetyl coenzyme A; PEP: phosphoenolpyruvate. Mean fold changes in transcription and p-values are indicated in Table 2.

Full size image

Temperature upshift is typically considered to be the main signal indicating to bacteria that they have entered the host; this hypothesis is supported by the thermal dependency of almost all Y. pseudotuberculosis virulence genes and also many of the latter's regulators [3]. Several of these genes were also found to be influenced by growth in plasma and the changes were sometimes in the opposite direction to those seen with temperature upshifts: whereas expression of the invasin-encoding gene inv was significantly repressed during bacterial growth under both conditions, transcription of psaA (coding for the pH6 antigen) was promoted by temperature upshifts [6, 20], but was one of the most strongly repressed in plasma. Interestingly, the impact of this medium on psaA transcription was not considered to be significant in Y. pestis and suggests that the pH6 antigen does not have the same importance in blood dissemination in the two species. In contrast to the latter two adhesins, transcriptional activation of yadA (harbored by the pYV plasmid and involved in adhesion) was found to be the highest of all the Y. pseudotuberculosis genes induced under plasma growth conditions. This observation is consistent with YadA's involvement in microbial resistance to complement [21, 22]. Similarly, ompC whose product is believed to be targeted by lactoferricin [23], a bactericidal peptide derived from lactoferrin by enzymatic cleavage [24], is strongly repressed, whilst no significant modification was observed for the outer membrane-encoding genes ompA and ompC 2.

Lastly, an essential determinant of bacterial virulence is the plasmid-encoded type III secretion system (TTSS) which performs intracellular delivery of a set of Yersinia outer proteins (Yops) that subvert the host's defenses [25]. Interestingly, Y. pseudotuberculosis growth in plasma induced the upregulation of 25 genes required for secretion, translocation and chaperoning of the Yop effector proteins in a similar fashion to that observed upon temperature upshift (Fig. 3). Furthermore, the apparently coordinated regulation of yadA and the TTSS-encoding genes by temperature and growth in plasma suggests the involvement of a common means of genetic control. YmoA (a chromatin-associated (histone-like) protein which is very similar in structure and function to the haemolysin expression modulating protein Hha from Escherichia coli) was shown to negatively influence YadA and Yop expression by favoring supercoiling of the pYV plasmid [26]. A two-fold reduction in ymoA transcription in plasma may be enough to contribute to the TTSS upregulation recorded in Y. pseudotuberculosis. Strikingly, this plasma-induced TTSS activation was not observed in Y. pestis, since only 3 out of the 25 genes mentioned above were found to be upregulated (in line with the statistically non-significant downregulation of ymoA); this raises the possibility that these two pathogenic Yersinia species may differ in their transcriptional regulation of pYV-harbored virulence genes.

Figure 3
figure 3

Medium- and temperature-dependent differential expression of genes harbored by the Y. pseudotuberculosis virulence plasmid pYV. Significant (p < 0.05) upshifts (yellow to red scale) or downshifts (blue scale) in individual gene transcription levels when bacteria were grown in human plasma versus LB (triangles) and/or at 37°C versus 28°C (squares) are indicated by the color scale bar. Only genes spotted on the macroarray (56 out of 99 pYV-borne genes) are shown and those encoding the secretion apparatus and Yop effectors are represented by grey and black boxes, respectively. Mean fold changes in transcription and p-values are indicated in Table 3.

Full size image

Conclusion

Overall transcription profiling of Y. pseudotuberculosis grown in an environment mimicking the blood stage of the infectious process revealed gene regulations that could not be anticipated from the results of previously reported single-stimulus studies. Our findings thus provide insight into how a number of simultaneously sensed environmental cues may be taken into account by the bacterium in a hierarchical manner. Furthermore, comparison of our analyses with those previously performed in Y. pestis suggests that transcription of common critical virulence factors may be differently influenced (at least in part) by the plasma environment in these two species.

Methods

DNA macroarray construction

Pairs of specific oligonucleotide primers were designed with the Primer 3 software for each of the 3,951 Y. pseudotuberculosis IP32953 CDSs. In order to avoid cross-hybridization, the specificity of the PCR products relative to the complete genome sequence was tested with CAAT-box software [27]. Primers purchased from Eurogentec were chosen in order to specifically amplify a ≈ 400 to 500 base pair (bp) fragment of each open reading frame (ORF), with a melting temperature of 51 to 60°C. Amplification reactions were performed in 96-well plates (Perkin-Elmer) in a 100 μl reaction volume containing 100 ng of Y. pseudotuberculosis IP32953 DNA, DNA polymerase (Dynazyme, New England Biolabs), 10 μM of each primer and 2 mM dNTPs (Perkin-Elmer). Reactions were cycled 45 times (94°C for 30 s; 60°C for 30 s; 72°C for 60 s) with a final cycle of 72°C for 7 min in a thermocycler. Each PCR product was checked by agarose gel electrophoresis and when DNA amplification was unsuccessful, PCR was repeated with another primer set. Overall, 3,951 of the 3,994 CDSs (98%) identified in the Y. pseudotuberculosis IP32953 genome were successfully amplified under our experimental conditions. ORF-specific PCR products, luciferase DNA (10 to 100 ng) and total genomic DNA from strain IP32953 were spotted onto 22 × 7-cm nylon membranes (Genetix) using a Qpix robot (Genetix). Immediately following spot deposition, membranes were immersed for 15 min in 0.5 M NaOH and 1.5 M NaCl, washed three times with distilled water and stored at -20°C until use. To ensure that DNA samples were successfully deposited on the membranes, 33P-labeled genomic DNA was hybridized to the macroarray before transcriptome analysis.

Bacterial culture

The Y. pseudotuberculosis transcriptome was studied in three independent cultures of strain IP32953 in media aliquoted from a single batch. After storage in Luria-Bertani (LB) broth with 40% glycerol at -80°C, the strain was thawed and then grown on LB agar supplemented with 20 μg ml-1 hemin for 48 h at 28°C. From this culture, 8 × 106 cells were inoculated into 40 ml of either LB broth or pooled human plasma from healthy donors (heated at 56°C for 30 min to ensure complement inactivation). Media were then incubated at 28°C or 37°C with shaking and Yersinia growth was monitored by absorbance at 600 nm.

RNA and cDNA probe preparation

Cells were harvested from exponential-phase cultures (A600 of 0.2–0.4 and 0.1–0.2 for LB and human plasma, respectively) by centrifugation at 4°C and the pelleted bacteria were disrupted with RNAwiz reagent (Ambion). After mixing the lysate with chloroform (0.2 v), total RNA was precipitated from the aqueous phase with glycogen (1/50 v) and isopropanol (1 v). The RNA pellet was washed with 70% ethanol and then dissolved in sterile, DNase- and RNase-free water. Contaminating DNA was removed using the DNA-free kit from Ambion. Nucleic acid purity and integrity was checked with a BioAnalyzer 2100 (Agilent) according to the supplier's instructions. After quantification by spectrophotometry at 260 and 280 nm, the RNA solution was stored at -80°C until use. cDNA was further generated from 10 μg of total RNA incubated (in a total volume reaction of 45 μl) for 3 h at 42°C with 50 U AMV reverse transcriptase (Roche), 0.35 pmol. of each amplified CDS-specific 3' oligonucleotide primer, 222 μM dATP, dGTP & dTTP, 2.2 μM dCTP and 50 μCi 33P-labelled dCTP (Amersham Biosciences). Labeled cDNA was purified to remove unincorporated nucleotides using DyeEx 2.0 spin column (Qiagen).

DNA macroarray hybridization

Macroarrays were prewetted in 2 × SSPE (0.18 M NaCl, 10 mM NaH2PO4, 1 mM EDTA, pH 7.7) and prehybridized for 1 h in 13 ml of hybridization solution (5 × SSPE, 2% SDS, 1× Denhardt's reagent, 0.1 mg of sheared salmon sperm DNA ml-1) at 65°C in roller bottles. Hybridization was carried out for 20 h at 65°C with 15 ml of hybridization solution containing the purified cDNA probe. After hybridization, membranes were washed three times at room temperature and three times at 65°C for 20 min in 0.5 × SSPE and 0.2% SDS. Probed macroarrays were exposed to a phosphor screen (Molecular Dynamics) for 24–72 h and imaged using a STORM 860 phosphorimager (Amersham Biosciences). The intensity of all of the pixels associated with each spot was further quantified using ArrayVision software (Imaging Research, Grinnel, IA, USA). The experiment design included three biological replicates for each combination of conditions. Data were analyzed using the SAS software (SAS Institute Inc, Cary, NC, USA). They were first log-transformed and normalized with a median normalization. A linear model was then applied on each gene with the temperature, phase and growth medium as fixed effects. The significance level alpha was set to 0.05.

Real-Time Quantitative PCR

Messenger RNAs (mRNAs) were reverse transcribed from 1 μg of nucleic acid by using the High-Capacity cDNA Archive Kit (Applied Biosystems, Foster City, CA) according to the manufacturer's instructions. The resulting cDNA was amplified by the SYBR Green Real-Time PCR Kit and detected on a Prism 7000 detection system (Applied Biosystems). The forward and reverse primers used were as follows: 5'CGCCATCAAATGCGCTAAT3' and 5'TGAGCGGGATCGTGTTCAA3' for yfeA, 5'TCAAGCAGGGAAACACATTCC3' and 5'GGCTGTTTACCCGCAAAAATC3' for psaA, 5'GGTTAGCCGCGAACAGGATA3' and 5'CGCTCGCCAGAACAAGGTT3' for aceB, 5'TCGATGCTCGCGCTAAGG3' and 5'GCTGGTTTCGCTGCTTCAG3' for yadA, 5'GATCCTGGTTCCATAAAAATTATTCAC3' and 5'ATTGTTCGCCTGGATTACCAA3' for yopJ, 5'GAGAATCCCAGTCGGGTGTTAA3' and 5'TCACTGCATCGCGGTAGGT3' for yopN, 5'GACACCAGTGGGACGCAACT3' and 5'GGGTTCACAAGAAAGAGTAACAGCTT3' for sycH, 5'GGTTACGCGCGGGTATCA3' and 5'CCGCGTCTTTGAGTGTTTTG3' for tnpR, 5'TTCTCGTGGGCAACCTATCC3' and 5'TGCGTTCCCAGCATACACAA3' for nlpD. On completion of the PCR amplification, a DNA melting curve analysis was performed to confirm the presence of a single amplicon. Relative mRNA levels (2ΔΔC) were determined by comparing the PCR cycle thresholds (Ct) for the gene of interest and the constitutively expressed YPTB0775 gene (spot ID YPO3356) coding for the outer membrane lipoprotein NlpD.