Evaluation of the Role of the opgGH Operon in Yersinia pseudotuberculosis and Its Deletion during the Emergence of Yersinia pestis

Abstract

The opgGH operon encodes glucosyltransferases that synthesize osmoregulated periplasmic glucans (OPGs) from UDP-glucose, using acyl carrier protein (ACP) as a cofactor. OPGs are required for motility, biofilm formation, and virulence in various bacteria. OpgH also sequesters FtsZ in order to regulate cell size according to nutrient availability. Yersinia pestis (the agent of flea-borne plague) lost the opgGH operon during its emergence from the enteropathogen Yersinia pseudotuberculosis. When expressed in OPG-negative strains of Escherichia coli and Dickeya dadantii, opgGH from Y. pseudotuberculosis restored OPGs synthesis, motility, and virulence. However, Y. pseudotuberculosis did not produce OPGs (i) under various growth conditions or (ii) when overexpressing its opgGH operon, its galUF operon (governing UDP-glucose), or the opgGH operon or Acp from E. coli. A ΔopgGH Y. pseudotuberculosis strain showed normal motility, biofilm formation, resistance to polymyxin and macrophages, and virulence but was smaller. Consistently, Y. pestis was smaller than Y. pseudotuberculosis when cultured at ≥37°C, except when the plague bacillus expressed opgGH. Y. pestis expressing opgGH grew normally in serum and within macrophages and was fully virulent in mice, suggesting that small cell size was not advantageous in the mammalian host. Lastly, Y. pestis expressing opgGH was able to infect Xenopsylla cheopis fleas normally. Our results suggest an evolutionary scenario whereby an ancestral Yersinia strain lost a factor required for OPG biosynthesis but kept opgGH (to regulate cell size). The opgGH operon was presumably then lost because OpgH-dependent cell size control became unnecessary.

INTRODUCTION

Yersinia pestis is the bacterium that causes plague, a fatal disease that cycles between mammalian and flea hosts (1). After Y. pestis is taken up into a flea's gut during a blood meal, the bacterium forms a biofilm that ultimately obstructs the digestive tract. The “blocked” (and thus starving) flea will bite a new host many times in an effort to feed. During these unproductive attempts to feed, some bacteria are dislodged from the biofilm and regurgitated into the dermal biting site (2–4). Although it is not clear whether these Y. pestis organisms are truly phagocytized, it generally assumed that they replicate initially within phagocytes and produce antiphagocytic factors; this leads to extracellular replication throughout the draining lymph node and ultimately in the blood and other deep tissues (5).

Y. pestis emerged from Yersinia pseudotuberculosis, a widely spread environmental bacterium that causes a mild bowel disease in humans following ingestion of contaminated foods (6). During its emergence, Y. pestis accreted a small amount of genetic material via horizontal transfer but also lost a large number of functional genes (7). Early investigations showed that the emergence of plague can be explained (at least in part) by the acquisition of genetic material (8–11). Hence, a stepwise scenario in which sequential gene losses led to a flea-borne transmission route was subsequently proposed (12, 13). However, the complete set of genetic events that led to the ability to cause plague remains unknown.

Genomic analysis has indicated that in comparison with all strains of Y. pseudotuberculosis for which the genome has been sequenced, all Y. pestis strains lack a 12-kb block encompassing 8 genes (including the opgGH locus) (see Fig. S1 in the supplemental material). The latter operon is found in many gammaproteobacteria and is functionally homologous to the nvbAB, chvAB, and cgs loci found in alphaproteobacteria (14–27). The opgGH operon encodes glucosyltransferases that synthesize branched glucans from UDP glucose (UDP-Glc), using an acyl carrier protein (ACP) as a cofactor (28, 29). In most bacterial species, the rate of glucan synthesis decreases in proportion to the osmolarity of the external microenvironment (20, 21, 25, 26, 28, 30–33). Accordingly, these macromolecules are generally referred to as osmoregulated periplasmic glucans (OPGs). It is noteworthy that Brucella and Rhizobiaceae may secrete OPGs into the external environment (15, 34).

Hence, in some (but not all) bacteria, OPGs may act as osmolytes that maintain turgor pressure during growth under low-osmolarity conditions (27). In gammaproteobacteria, OPGs reportedly control not only cell motility, biofilm formation, and virulence (via the regulation of the Rcs regulatory system) but also protein folding and degradation and carbohydrate catabolism (via as-yet-unknown mechanisms) (16, 24, 35–40). Furthermore, OPGs can sequester antibiotics and detergents, thus protecting the bacterium against these harmful molecules (40–42). When secreted in the environment, OPGs were found to prevent phagolysosome biogenesis and thus enable intracellular replication of Brucella (34, 43). OPGs are also involved in nodule production in plants that host symbiotic bacteria (22, 23, 32, 44, 45). Hence, in all bacterial species studied to date, OPGs appear to be key players in virulence and in bacterial adaptation to environmental change.

Lastly, OpgH controls bacterial cell size and shape in an UDP-Glc concentration-dependent manner (independently of its glucosidase activity and thus its effect on the Rcs system) (46). In response to an increase in the intracellular concentration of UDP-Glc (i.e., passage into a nutrient-rich environment), OpgH reduces the rate of the cytokinetic ring formation by directly sequestering the cell division protein FtsZ at midcell. This results in a greater cell size but does not affect the mass doubling time. Hence, OpgH's sequestering activity might provide a selective advantage in response to changes in nutrient availability, because bacterial cell shape is considered to have selective value (47). However, this idea remains to be proven.

Given this context, the absence of opgGH in Y. pestis is intriguing because the bacterium is much more virulent than Y. pseudotuberculosis and forms biofilms for efficient flea-borne transmission (8). Hence, we sought to establish why Y. pestis has lost the opgGH operon and whether this loss contributed to the emergence of modern plague. However, we first had to study opgGH's previously unknown role in Y. pseudotuberculosis.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

The strains and plasmids used in the present study are listed in Table S1 in the supplemental material. Lysogeny broth (LB) with or without sodium chloride (NaCl), low-osmolarity medium (LOS) (21), and LB agar with or without kanamycin (50 μg/ml) or carbenicillin (100 μg/ml) supplementation were used to culture bacteria at the desired temperature. opgGH-negative Y. pseudotuberculosis mutants (in which bases 1070 to 1829 are deleted) were generated using the pCVD442-based method and checked in a PCR analysis, as previously described (48). The opgGH and galUF operon from Y. pseudotuberculosis and the acpP gene from Escherichia coli (under the control of their respective promoters) were amplified by PCR using the primer sets Opg1 (5′-CGCGCTGCAGCTGACTTGCTAGGCTTATGC-3′) plus Opg2 (5′-GTGTGGATCCCTCTTATGGTCTGCCGCTAC-3′); Gal1 (5′-AAAAGAGCTCAGATGTTGGATATTATTTTCA-3′) plus Gal2 (5′-ACATGCATGCACCGCCATACCGACAACGCCGA-3′); and Acp1 (5′-CCGCGGATCCTGATTTGCGTTATTGGGGGG-3′) plus Acp2; (5′-CCCGGAATTCTAAAACTCAGGCGGTCGAAC-3′). The opgGH and galUF amplicons were digested with the restriction enzymes BamHI/PstI and EcoRI/SphI, respectively, and were thus inserted into the corresponding restriction sites of the pUC18Not and pBR322 vectors (yielding the plasmids pNF400 and pRB100). The acpP amplicon was cloned into the pCRII cloning vector (Invitrogen) to yield pAcpEc. Lastly, pNF400 was used to subclone the Y. pseudotuberculosis opgGH operon at the BamHI and PstI sites in the mini-Tn7 pUC842 vector and thus integrate the operon into the Y. pestis chromosomal at att Tn7 (49).

Extraction of OPGs.

OPGs were extracted as previously described (28). Briefly, bacteria cultured under the chosen growth conditions were pelleted by centrifugation, suspended in sterile, distilled water, and then lysed with trichloroacetic acid. After centrifugation of the lysate, the supernatant was collected and mixed with charcoal. After incubation at room temperature with vigorous shaking, the charcoal (which bound the OPGs) was collected by centrifugation and mixed with pyridine to release the glucans. After centrifugation, the OPG-containing supernatant was concentrated by rotary evaporation and then fractionated by gel filtration. The presence of OPGs in the eluate fractions was monitored using the anthrone-sulfuric acid method (50). The putative secretion of OPGs into the growth media was determined by applying the above-described procedure to the culture supernatant.

Motility assay.

Bacteria grown overnight in LB at 21°C (for Y. pseudotuberculosis) or 30°C (for E. coli and Dickeya dadantii [previously Erwinia chrysanthemi]) were diluted in LB to an optical density at 540 nm (OD₅₄₀) of 0.2. Five microliters of the freshly diluted suspension was spotted at the center of an LB soft agar plate containing 0.25% agar (for Y. pseudotuberculosis) or 0.4% agar (for E. coli and D. dadantii). Bacterial swimming diameters were measured after a 24-hour incubation at 21°C (for Y. pseudotuberculosis) or at 30°C (for E. coli and D. dadantii).

Biofilm assay.

Biofilms were assayed as previously described, with slight modifications (12). Briefly, bacteria suspended in LB supplemented with 4 mM MgCl₂ and 4 mM CaCl₂ was added to the wells of a 24-well plate. After a 24-hour incubation at 21°C with shaking, the planktonic bacteria were rinsed away with water and the attached bacterial biofilm was stained with a crystal violet dye solution. After staining, the wells were washed with water. The dye attached to the biofilm was then released by the addition of an ethanol-acetone solution. The absorbance of each well was measured at 540 nm. Measurements were corrected by subtracting the crystal violet binding observed for control (noninoculated) wells. The absorbance ratio for each strain of interest (relative to the parental strain) was calculated.

Analysis of opgGH expression.

In transcription assays, cultures were mixed with RNAprotect reagent (Qiagen) according to the manufacturer's protocol and then pelleted for RNA extraction with the Nucleospin RNA kit (Macherey-Nagel). RNA samples were (i) reverse transcribed using the Superscript III kit (Invitrogen) after having been treated twice with DNase (using a DNA-free kit from Ambion), (ii) checked for integrity (using an Agilent Bioanalyzer), and (iii) checked for the absence of DNA contamination (using PCR). Quantitative PCR was performed using SYBR green technology with the primer sets opg3 (5′-GCTGACGGTAATAGCCATTCCAT-3′) plus opg4 (5′-CCCGTTTGCATCAGCAGATT-3′) and crr1 (5′-GCCCTCTGGCAATAAA ATGG-3′) plus crr2 (5′-AGCATGGTTGGTCTCGAAAATT-3′). Data were normalized against levels of the crr transcript, and the relative fold change was calculated using the ΔΔC_T method.

In translation assays, whole-cell lysates were separated by sodium dodecyl sulfate-polyacrylamide (10%) gel electrophoresis and then electrophoretically transferred to a nitrocellulose membrane using equipment and protocols from Bio-Rad Laboratories. The membrane was blocked overnight at 4°C with 5% (wt/vol) nonfat dry milk in phosphate-buffered saline containing 0.2% Tween 20 (PBS-T), washed three times for 10 min with PBS-T, probed at room temperature for 1.5 h with a rabbit polyclonal against His-tagged OpgG from E. coli (Eurogentec) diluted in PBS-T, and then washed three times with PBS. Immunoreactive proteins were detected using a horseradish peroxidase-conjugated goat anti-rabbit IgG and the ECL plus Western blotting kit (GE Healthcare), according to the manufacturer's protocols.

Measurement of bacterial cell size.

Smears of mid-log-phase bacterial cultures in LB supplemented with glucose (2 g/liter) were air dried and ethanol fixed on glass slides. Fixed bacteria were successively washed with 70% ethanol, stained with fuchsin, and then washed with water. The cell length (250 cells per slide) was measured using ImageJ software.

Interactions with serum, polymyxin, and macrophages.

Survival and growth upon contact with serum, the MIC of polymyxin, and survival within macrophages were assayed as previously described (51). Briefly, a bacterial suspension of 5 × 10⁸ bacteria/ml in PBS was prepared from an overnight culture in LB at 21°C. The suspension was diluted 10-fold into fresh human serum and then incubated at 37°C. Bacterial growth was measured by counting the CFU. To determine the MIC of polymyxin, a range of polymyxin concentrations were added to inocula of 10⁶ bacteria/ml (prepared from an overnight culture at 21°C or 37°C in LB). The MIC was measured after a 24-hour incubation. For the assay of survival within phagocytes, bacteria from an overnight culture at 21°C were suspended in Dulbecco's modified Eagle's medium (DMEM) medium and added to RAW macrophages suspended in DMEM at multiplicity of infection of 10. Cells were centrifuged for 5 min at 18 × g and then incubated at 37°C in a 5% CO₂ atmosphere. Thirty minutes after contact, nonphagocytized bacteria were rinsed away with DMEM, and fresh DMEM supplemented with gentamicin was then added. At different time points, the supernatant was removed, cells were lysed with cold water on ice, and serial dilutions of the lysate were spread on blood agar to count CFU.

Determination of bacterial virulence.

Groups of six to eight 9-week-old OF-1 female mice (Charles River) were inoculated (i) via intragastric and intravenous routes with Y. pseudotuberculosis strain 2777 or its isogenic opgGH mutant or (ii) via the intradermal route with Y. pestis CO92 strain or its derivative, as previously described (5, 52). Mice were monitored daily after inoculation. To measure the time course of organ colonization, organs (Peyer's patches, mesenteric lymph nodes, spleens, and livers) were collected in PBS immediately after euthanasia and put on ice. Dilutions of the triturated organs were plated on LB agar for CFU counting after 48 h of incubation at 28°C. The virulence of the phytopathogen D. dadantii on chicory leaves was determined as previously described (24). Briefly, a bacterial suspension in saline solution was prepared from an overnight culture in LB medium and placed on a wounded leaf. Maceration was determined after 48 h of incubation in a dew chamber at 30°C. At the time when the present experiments were initiated, they were not subject to ethical approval. However, the legislation changed during the study. Accordingly, some (but not all) of the experiments were approved by the regional animal care and use committee (Lille, France). However, all animal experiments were performed in compliance with French and European regulations on the care and protection of laboratory animals (EC Directive 86/609 and the French Act 2001-486, issued on 6 June 2001).

Flea infection.

Y. pestis strain KIM6+ and its derivative were cultured in brain heart infusion (BHI) for 18 h at 37°C (without shaking), pelleted by centrifugation, and then suspended to 3 × 10⁸ bacteria/ml in PBS. Next, 1 ml of the suspension was homogenized with 5 ml of heparinized mouse blood. The contaminated blood was added to an artificial feeding system in which ∼400 Xenopsylla cheopis fleas were allowed to feed for 1 h. After feeding, two cohorts of infected fleas were collected, maintained at 21°C and 75% humidity, and allowed to feed twice a week for a 4-week period. Notably, one cohort (composed of 50 males and 50 females) was used to monitor blockage of the flea gut. The second cohort (composed of females only) was used to measure infection rates and bacterial loads immediately after and 27 days after the infectious blood meal. In particular, 20 insects were individually triturated and plated on BHI containing 1 μg/ml Irgasan and 10 μg/ml hemin. CFU were counted after a 48-hour incubation at 28°C.

Statistical analysis.

A one-way analysis of variance, the Gehan-Breslow-Wilcoxon test, and the Mann-Whitney U test were used to analyze the results of in vitro experiments, survival curves, and data from flea infection experiments.

RESULTS

In contrast to other proteobacteria, Y. pseudotuberculosis does not produce detectable OPGs.

The role of opgGH in Y. pseudotuberculosis was not known prior to the present study. Hence, we first had to determine its role in order to establish whether loss of opgGH was likely to have been selected during the emergence of plague or, in contrast, had no impact. Surprisingly, no OPGs could be detected in Y. pseudotuberculosis strain 2777 grown under low-osmolarity conditions (NaCl-free LB and LOS) when the assay method conventionally employed for all bacterial species studied to date (including E. coli and D. dadantii, which were used as positive controls here) was used (Fig. 1A and data not shown). No OPGs were detected in cultures of Y. pseudotuberculosis strain 2777 or in the extracellular medium when the extraction volume was increased 10-fold (i.e., up to 1 liter of culture medium) (data not shown). These results were not due to an artifact with strain 2777, because we did not detect any OPGs for the unrelated Y. pseudotuberculosis strain 2515. However, Western blot experiments indicated that Y. pseudotuberculosis does translate the opgGH operon (Fig. 2B).

FIG 1.

Y. pseudotuberculosis does not produce detectable OPGs, despite the fact that its opgGH operon encodes functional active glucosyltransferases in E. coli and D. dadantii. (A) Eluate fractions of bacterial samples (prepared for OPG extraction, as described in Materials and Methods) stained with anthrone-sulfuric acid reagent. Blue color indicates the presence of OPGs. In these experiments, WT Y. pseudotuberculosis, E. coli, and D. dadantii strains and their derivative opgGH- or opgG-negative strains harboring (or not) the Y. pseudotuberculosis opgGH operon and its own promoter on a plasmid (pNF400; +opgGH_Ypst) were grown in NaCl-free LB. (B) Swimming motility of WT E. coli and D. dadantii strains (black bars) and their OPG-negative derivative strains harboring pNF400 (+opgGH_Ypst) (gray bars) or not (white bars), after a 24-hour incubation on a LB soft agar plate at 30°C. The photos are representative of bacteria after 24 h of swimming, and the swim diameter (mean ± SEM) was determined from three independent experiments. The swim diameters of the ΔopgGH E. coli strain and the ΔopgG D. dadantii strain were significantly lower than those of the corresponding WT strain and the respective OPG-deficient mutant strain expressing opgGH from Y. pseudotuberculosis (P < 0.05 in a one-way analysis of variance). However, partial complementation was observed in D. dadantii, presumably due to plasmid instability in this species. (C) Virulence of D. dadantii on chicory leaves after 48 h of contact. Leaves were infected with 10⁷ CFU of WT D. dadantii, ΔopgG mutant, or ΔopgG mutant harboring pNF400 (+opgGH_Ypst). *, P < 0.05.

FIG 2.

Y. pseudotuberculosis constitutively expresses the opgGH operon under various growth conditions. (A) Fold change expression of opgGH in Y. pseudotuberculosis (strain 2777) grown under various conditions and in Y. pseudotuberculosis harboring a recombinant pBR322 plasmid containing its galUF operon or a recombinant pUC18 plasmid containing its opgGH operon, compared with the opgGH expression level in Y. pseudotuberculosis grown under low-osmolarity and aerobic conditions and at 28°C. galUF and opgGH were under the control of their own respective promoters. Under each condition, the level of opgGH transcripts was normalized to that of crr transcripts. The data are means and SEM from two independent experiments. (B) Immunoblot of a whole-cell lysate of Y. pseudotuberculosis (strain 2777), its isogenic ΔopgGH mutant and WT E. coli grown in LB at 30°C. E. coli was used as a positive control. The blot was immunostained with a polyclonal antibody against E. coli OpgG. Arrowheads indicate the positions of the Y. pseudotuberculosis OpgG protein.

The opgGH operon from Y. pseudotuberculosis encodes functional OPGs in E. coli and D. dadantii.

The above-mentioned, unexpected results prompted us to determine whether the opgGH operon from Y. pseudotuberculosis encodes functional enzymes. To this end, we looked at whether opgGH from Y. pseudotuberculosis restores the phenotypes associated with the loss of OPG synthesis in E. coli and D. dadantii. In the absence of OPGs, both E. coli and D. dadantii show motility defects, and D. dadantii is also avirulent on chicory leaves (24, 38). OPGs were purified from OPG-deficient strains (grown in NaCl-free LB) of E. coli (opgG-negative, opgH-negative, and opgGH-negative strains) and D. dadantii (an opgG-negative strain), in which the opgGH operon from Y. pseudotuberculosis was expressed under the control of its own promoter (Fig. 1A and data not shown). We found that motility was restored in both E. coli and D. dadantii opgG-mutant strains and that D. dadantii was newly virulent on chicory leaves (Fig. 1B and C).

Y. pseudotuberculosis constitutively transcribes opgGH but does not produce detectable OPGs under various growth conditions.

Given that opgGH from Y. pseudotuberculosis does indeed encode functional enzymes, we next sought to identify growth conditions that could stimulate OPG production, i.e., whether synthesis of OPGs is regulated at transcriptional and/or posttranscriptional level in Y. pseudotuberculosis. To this end, we monitored (i) the transcription of opgGH using quantitative reverse transcription-PCR (qRT-PCR) and (ii) levels of the OpgGH glucosyltransferase product under various growth conditions (high-, low-, and medium-osmolarity media; 21°C and 37°C; aerobic and anaerobic conditions; and heat-inactivated serum [which was used here as a surrogate model of in vivo conditions]). The assay results indicated that opgGH was constitutively transcribed under all the conditions tested (Fig. 2A) but that no OPGs were detected.

Overexpression of opgGH by Y. pseudotuberculosis does not induce OPG biosynthesis.

The above data suggested that Y. pseudotuberculosis lacks essential OPG synthesis factors and/or carries genes encoding inhibitors of OPG biosynthesis. To distinguish between these two possibilities, we decided to screen for OPGs in Y. pseudotuberculosis overexpressing opgGH (since genetic overexpression can override inhibition mechanisms). We therefore transformed Y. pseudotuberculosis with a high-copy-number plasmid (pUC18) containing opgGH (under the control of its own promoter), confirmed (using qRT-PCR) that the operon is highly transcribed, and then assayed the transformed strain for OPG synthesis. Even when the recombinant strain overexpressed opgGH 100-fold more intensely than the parental strain, OPGs were still not detected (Fig. 2A and data not shown). Lastly, OPG synthesis was not detected when Y. pseudotuberculosis overexpressed opgGH from E. coli (under the control of its own promoter); the data were identical to those shown in Fig. 1A. This finding suggests that Y. pseudotuberculosis lacks factors that are important for OPG biosynthesis.

The absence of detectable OPGs in Y. pseudotuberculosis is not due to a lack of functional ACP.

The ACP is essential for OPG synthesis in E. coli (29), and certain amino acid substitutions can affect the protein's ability to participate in OPG biosynthesis (14). Hence, our inability to detect any OPGs in Y. pseudotuberculosis might have been caused by a nonfunctional ACP. Consistently, an amino acid sequence alignment revealed that ACP from Y. pseudotuberculosis and E. coli diverge (see Fig. S2 in the supplemental material). However, we did not detect any OPGs in Y. pseudotuberculosis expressing acpP (under the control of its own promoter) from E. coli (data not shown). Hence, ACP does not appear to account for the lack of OPGs.

The absence of detectable OPGs in Y. pseudotuberculosis is not due to a defect in UDP-Glc synthesis.

UDP-glucose is an essential precursor for the biosynthesis of several polysaccharides, including OPGs (27). Hence, one can hypothesize that Y. pseudotuberculosis does not produce enough UDP-glucose to drive the production of OPG. To test this hypothesis, the galUF operon (governing UDP-Glc synthesis [53]) was overexpressed (using the average-copy-number plasmid pBR322) in Y. pseudotuberculosis. This overexpression strongly induced the transcription of the opgGH operon but did not lead to the production of detectable OPGs (Fig. 2A and data not shown), further supporting the hypothetical lack of essential factors for OPG biosynthesis.

The loss of opgGH does not impact Rcs-dependent phenotypes in Y. pseudotuberculosis.

The above data suggested that Y. pseudotuberculosis does not synthesize OPGs. However, we could not rule out the presence of very small amounts of extremely active OPGs (or other glucans) that might nevertheless control the Rcs regulatory system. Hence, we next looked at whether a ΔopgGH Y. pseudotuberculosis strain displayed one or more phenotypes (defects in motility, biofilm formation, and virulence) attributed to aberrant stimulation of the Rcs system by OPGs (24, 36, 38). As shown in Fig. 3A and B, ΔopgGH Y. pseudotuberculosis remained fully motile and formed the same amount of biofilm as the wild-type (WT) strain. Furthermore, the mutant grew normally within macrophages (Fig. 3C). The MICs of polymyxin for the mutant grown at 21 and 37°C were identical to those observed for the parental strain (48 μg/ml and 192 μg/ml, respectively), suggesting that the absence of opgGH did not impact virulence. Consistently, the time course of organ colonization in animals inoculated by intragastric and intravenous routes revealed that the mutant colonized the Peyer's patches, the mesenteric lymph nodes, the spleen, and the liver at much the same rate as the WT strain (Fig. 3D and E).

FIG 3.

In contrast to other proteobacteria, Y. pseudotuberculosis lacking a functional opgGH operon does not show the pleiotropic phenotype resulting from the activity of the OPGs through the Rcs signaling pathway. Data on swimming motility on an LB soft agar plate after a 24-hour incubation at 21°C (A), biofilm formation (relative to the WT strain) after a 24-hour incubation at 21°C with shaking in LB supplemented with Ca²⁺ and Mg²⁺ (B), rates of survival within RAW 264.7 macrophages up to 10 h after internalization (C), the time course of colonization of the Peyer's patches and mesenteric lymph nodes after intragastric inoculation (10⁹ bacteria) (D), and the time course of colonization of the spleen and liver after intravenous inoculation (10³ bacteria) by the WT strain 2777 and its derivative ΔopgGH strain (E) are shown. The photos are representative of bacteria after 24 h of swimming. The means and SEM from three independent experiments (A and B) and two independent experiments (C) are shown. The time course of organ colonization was determined from one experiment using groups of six animals (D and E); horizontal lines indicate the medians of the individual data points. Bacterial loads in organs did not vary significantly according to the presence or absence of opgGH (P > 0.05 in a one-way analysis of variance).

Loss of opgGH during emergence probably explains why Y. pestis is smaller than Y. pseudotuberculosis.

Independently of its ability to synthesize OPGs, OpgH is involved in elaboration of the bacterial cell's architecture by sequestering FtsZ and thus affecting the cell length (46). Hence, we looked at whether the loss of opgH in Y. pseudotuberculosis affects bacterial shape during growth in LB supplemented with glucose and at the temperatures encountered by the bacterium in nature (21°C and 37°C). Regardless of the growth temperature, the opgGH mutant was significantly (11% to 14%) smaller than the WT strain (Fig. 4). However, the cell size was normal when the mutant was complemented with the WT opgGH operon. Consistently, Y. pestis (which lost opgGH during emergence) was significantly (∼11%) smaller than Y. pseudotuberculosis when cultured at 37°C. Furthermore, expression of the Y. pseudotuberculosis opgGH operon from the chromosomally integrated mini-Tn7 increased the length of Y. pestis to the same size as its recent ancestor (1.64 μm and 1.66 μm, respectively). Surprisingly, Y. pestis grown at 21°C was the same size as Y. pseudotuberculosis and was greater still (by 8%) when the bacillus expressed opgGH from Y. pseudotuberculosis. However, in a one-way analysis of variance, the latter difference was not statistically significant. Taken as a whole, and considering that OpgH (but not OpgG) was found to control cell size in E. coli (46), these data suggest that (i) opgH controls the length of Y. pseudotuberculosis and (ii) deletion of opgH during the emergence of Y. pestis decreased the bacterial cell size at the mammalian host temperature but not at a temperature required for efficient flea-borne transmission.

FIG 4.

Y. pestis is smaller than Y. pseudotuberculosis when grown at 37°C (but not at 21°C), presumably because the bacterium lost opgGH during its emergence. The lengths of Y. pseudotuberculosis WT strain 2777, the ΔopgGH strain, the complemented mutant, and Y. pestis KIM6+ expressing opgGH from Y. pseudotuberculosis (+opgGH) or not (WT) were measured during exponential-phase growth in LB supplemented with glucose at 21°C and at 37°C. The data are means and SEM from five independent experiments. Regardless of the growth temperature, the ΔopgGH Y. pseudotuberculosis strain (but not the complemented mutant strain) was significantly smaller than the WT strain (P < 0.05 in one way analysis of variance). When grown at 37°C, Y. pestis was significantly smaller than Y. pseudotuberculosis except when it expressed opgGH from Y. pseudotuberculosis (P < 0.05 in one way analysis of variance). *, P < 0.05.

Effect of the expression of opgGH from Y. pseudotuberculosis on the Y. pestis life cycle.

A particular bacterial shape is thought to have selective value by facilitating specific cell functions (47). Hence, we wondered whether the loss of opgGH provided Y. pestis with a selective advantage for the emergence of flea-borne plague. Virulence of Y. pestis containing a copy of opgGH at the att Tn7 chromosomal site was assessed in a mouse model of bubonic plague and in the Xenopsylla cheopis rat flea model. Regardless of the presence or absence of opgGH in the Y. pestis genome, all mice inoculated with ∼10 CFU succumbed to the disease. A slight increase in the median survival time was observed for animals inoculated with Y. pestis expressing opgGH (4 days, versus 3 in Y. pestis not expressing opgGH), although the survival curves were not significantly different (P > 0.3). Consistently, we found that expression of opgGH did not affect the bacterium's capability to survive within macrophages or to grow in normal human serum (see Fig. S3 in the supplemental material). Furthermore, expression of opgGH did not affect the bacillus' capability to block the flea digestive tract (Fig. 5A). Accordingly, 27 days after the infected blood meal, the percentage of infected fleas and the bacterial loads recovered from the infected fleas did not differ according to the expression or nonexpression of opgGH (Fig. 5B and C). These data are line with the fact that biofilm formation in vitro was not affected in Y. pestis expressing opgGH (biofilm formation, reported as the mean ± standard error of the mean [SEM], was 89 ± 9.6% relative to the parental strain, as determined using data from two independent experiments).

FIG 5.

Y. pestis expressing opgGH from Y. pseudotuberculosis is fully competent in fleas. (A) Proportion of fleas became blocked in the 4 weeks following an infected blood meal; (B) proportion of infected fleas; (C) bacterial loads immediately after and 27 days after the infectious meal. Experiments were performed with Y. pestis WT KIM6+ (WT) and Y. pestis KIM6+ harboring a chromosomally integrated mini-Tn7 encompassing the opgGH operon from Y. pseudotuberculosis (+opgGH). Data in panels A and B are means and standard deviations (SD) from two independent experiments. In panel C, the data correspond to samples from two independent experiments. Each circle indicates the bacterial load of an individual flea, and the horizontal lines indicate the median CFU per flea.

DISCUSSION

All of the 20 or so bacterial species studied to date that carry the opgGH operon or one of its functional homologs (ndvAB, chvAB, and cgs) have been found to govern the biosynthesis of the glucose homopolymer OPG. The latter accounts for 0.5% to 10% of the bacterial dry weight (22–24, 32, 34, 36, 40, 44, 45). However, using the conventional method used to extract OPGs from all the bacterial species studied to date, we did not detect any OPGs (i) in Y. pseudotuberculosis or its culture supernatant after bacterial growth under various growth conditions, (ii) when the bacterium artificially overexpressed its own opgGH operon or that of E. coli, (iii) when Y. pseudotuberculosis overproduced the Opg enzymes' substrate (UDP-Glc) via the overexpression of galUF, or (iv) when the bacterium expressed the acpP gene from E. coli, which encodes a cofactor required for OPG biosynthesis. Lastly, several cellular processes (motility, biofilm formation, and virulence) known to be controlled by OPGs via the Rcs signaling pathway in E. coli and/or D. dadantii (24, 38) were unaffected in a ΔopgGH Y. pseudotuberculosis strain. However, the Rcs system is known to control biofilm formation and motility in Y. pseudotuberculosis (54). Taken as a whole, these data suggest that Y. pseudotuberculosis has lost the capability to produce OPGs.

One can legitimately wonder why Y. pseudotuberculosis (probably) does not produce OPGs, considering that (i) our present data and data from the KEGG database suggest that the bacterium produces the Opg enzymes, the ACP cofactor, and the UDP-Glc substrate (following glucose incorporation or glucogenesis) and (ii) the opgGH operon from Y. pseudotuberculosis encodes functional glucans in E. coli and D. dadantii. On one hand, Y. pseudotuberculosis might produce OPGs under particular conditions or might lack an essential factor for OPG biosynthesis. The latter might be a polyprenyl-phosphate carrier because this conserved molecule has a key role in glycan biosynthesis in several kingdoms, and it has been suggested that UPD-Glc has to be transferred to a polyprenyl-phosphate carrier prior to the assembly of OPGs (55, 56). On the other hand, Y. pseudotuberculosis may constitutively accumulate certain ionic and/or nonionic solutes potentially involved in controlling glycosyltransferase activity in response to osmotic stress (57, 58). However, we found that 100-fold overexpression of the opgGH operon did not bypass this putative repression. Furthermore, the screening of ∼1,000 strains from a Tn5 mutant library of Y. pseudotuberculosis for OPG synthesis did not lead to the identification of this type of inhibitor (unpublished data). Hence, the data support the hypothesis whereby Y. pseudotuberculosis lacks a functional intermediate that is essential for OPG biosynthesis.

The apparent absence of OPGs in both Y. pseudotuberculosis and Y. pestis (the latter being naturally devoid of opgGH) raises the question of how these species compensate for the lack of OPGs. Suppressor mutations that rescue phenotypes associated with loss of OPGs have been identified in E. coli, D. dadantii, and members of the Rhizobiaceae (35–38, 59–61). All of these mutations affect regulatory systems, including OmpR-EnvZ, RcsCDB, and a phosphodiesterase regulating cyclic di-GMP signaling. However, none of the suppressor mutations completely alleviates the loss of OPGs. Hence, Y. pseudotuberculosis and Y. pestis may have several suppressor mutations, i.e., a very different regulatory network compared with the above-mentioned bacteria. Consistently, Y. pestis has altered Rcs and cyclic di-GMP signaling pathways and does not control ompF and ompC (involved in bacterial adaptation to osmotic conditions) in the same way as E. coli and D. dadantii (13, 62, 63). However, whatever applies to Y. pestis might not apply to Y. pseudotuberculosis, since the latter has a functional Rcs system and does not control its c-di-GMP signaling pathways in the same way as the former. Differences in the membrane composition in Y. pseudotuberculosis and Y. pestis (relative to E. coli and D. dadantii) may constitute an alternative and/or additional explanation (64). Indeed, membrane composition has a role in bacterial adaptation to the environment (65, 66). Lastly, Y. pseudotuberculosis and Y. pestis may produce glucans that are equivalent to OPGs, such as the “free oligosaccharides” described in Campylobacter jejuni (67).

In agreement with a previous report on E. coli (46), we found that OpgH controls the size of Y. pseudotuberculosis. Consistently, Y. pestis (which naturally lacks opgGH) is smaller than Y. pseudotuberculosis but was the same size as its recent ancestor when it expressed opgGH. However, Y. pestis and Y. pseudotuberculosis differed in size when cultured at 37°C but not when cultured at 21°C, presumably because Y. pestis has a compensatory mechanism. These results indicate that the shape control mechanism has evolved in Y. pestis, potentially providing a selective advantage. Indeed, it has been found that a particular cell shape can maximize the use of available nutrients and promote biofilm formation in vitro and virulence (e.g., the ability to escape detection and resist phagocytes and complement) (68–73). However, opgH expression did not influence the virulence of Y. pestis in mice or its ability to colonize and block fleas. In agreement with the virulence data for Y. pestis, Y. pseudotuberculosis lacking OpgH was able to colonize mice normally.

Lastly, our results suggest an evolutionary scenario in which an ancestral strain of Yersinia lost (or did not acquire) a cofactor that was important for OPG biosynthesis. However, opgGH was initially retained, presumably because the ability to control bacterial size provides an advantage in the external environment. We suggest that the opgGH gene was subsequently lost because the OpgH-dependent size control mechanism was no longer necessary or was replaced by another control mechanism. During this evolutionary process, several events (including regulatory network reprogramming) presumably occurred so as to rescue the essential cellular functions associated with OPGs synthesis and the OpgH-dependent size-control mechanism identified in other bacteria. However, there are other possible explanations for the loss of opgGH. Indeed, we cannot rule out the possibility that Y. pseudotuberculosis synthesizes OPGs in order to survive under specific conditions that have not been identified in other bacteria. Consistently, genomic analysis indicates that Y. enterocolitica (the third human-pathogenic Yersinia sp.) and all environmental Yersinia species (for which the genomes have been sequenced) have an opgGH operon. Hence, Y. pestis may also have lost the opgGH operon because it no longer encounters conditions under which OPG synthesis is required for survival.