Pulmonary infection by Yersinia pestis rapidly establishes a permissive environment for microbial proliferation

Abstract

Disease progression of primary pneumonic plague is biphasic, consisting of a preinflammatory and a proinflammatory phase. During the long preinflammatory phase, bacteria replicate to high levels, seemingly uninhibited by normal pulmonary defenses. In a coinfection model of pneumonic plague, it appears that Yersinia pestis quickly creates a localized, dominant anti-inflammatory state that allows for the survival and rapid growth of both itself and normally avirulent organisms. Yersinia pseudotuberculosis, the relatively recent progenitor of Y. pestis, shows no similar trans-complementation effect, which is unprecedented among other respiratory pathogens. We demonstrate that the effectors secreted by the Ysc type III secretion system are necessary but not sufficient to mediate this apparent immunosuppression. Even an unbiased negative selection screen using a vast pool of Y. pestis mutants revealed no selection against any known virulence genes, demonstrating the transformation of the lung from a highly restrictive to a generally permissive environment during the preinflammatory phase of pneumonic plague.

Most lung pathogens induce a measureable inflammatory response within a few hours of infection, but Yersinia pestis is strikingly different: the first 36 h of infection are characterized by rapid bacterial replication in the lung without generating appreciable levels of proinflammatory cytokines/chemokines, histological changes in infected tissue, or outward disease symptoms. This process allows the organisms to multiply rapidly before the inflammatory response is fully activated, typically too late to control the infection. Thus, disease progression for primary pneumonic plague can be subdivided into two distinct phases: an initial preinflammatory phase followed by a proinflammatory phase, which is recapitulated in both mouse and nonhuman primate models of infection (1–4). The absence of early pulmonary inflammation in response to Y. pestis could be because of the organisms avoiding detection by the immune system, a strategy used by many pathogens, including Y. pestis (5–7); or, it could be because of the organisms completely suppressing early innate immune responses, which has never been observed for bacterial pathogens.

The importance of an early preinflammatory phase is demonstrated by at least two previous studies. First, Y. pestis exhibits decreased replication in the lungs of mice with a heightened or altered state of immune activation induced by a latent herpesvirus infection (8), suggesting that preexisting or early inflammatory responses are able to partially control and mitigate disease progression. Second, a Y. pestis mutant lacking YopH, an effector protein of the Yersinia Ysc type III secretion system (T3SS), is immunostimulatory and also growth-inhibited during respiratory infection (9). Based on these studies, we hypothesized that the ability of Y. pestis to actively suppress early innate immune mechanisms may have a more significant role in primary pneumonic plague than any intrinsic nonstimulatory properties of the bacterium itself. We evaluated the early consequences of Y. pestis respiratory infection using a coinfection model of pneumonic plague and discovered that the organisms are responsible for a rapid and profound modulation of pulmonary defenses that allows, not only these organisms to proliferate rapidly in the lung, but other microbes as well.

Results and Discussion

Pulmonary trans-Complementation of Avirulent Mutants by Wild-Type Y. pestis.

We designed a series of in vivo coinfection experiments to test the idea that Y. pestis produces a dominant immunosuppressive state in the lung. Using a genetically marked wild-type strain [ΔlacZ mutant of Y. pestis CO92 (10) -YP160] (Fig. S1) to quantitatively measure competition, we coinfected C57BL/6J mice with a Y. pestis ΔyopH strain to see if the immunostimulatory response to the mutant could be “suppressed” by the wild-type bacteria. After intranasal inoculation of a 1:1 mixture of wild-type and ΔyopH strains, we observed that wild-type bacteria were able to almost fully restore the in vivo replication defect of a coinfecting ΔyopH mutant (Fig. 1A).

Fig. 1.

Wild-type Y. pestis is able trans-complement avirulent isogenic strains of Y. pestis. (A) Individual intranasal infections (white bars) for ΔyopH (YP373), Δpla (YP102), or pCD1⁻ (YP6) mutants or matched coinfections (gray bars) with wild-type Y. pestis in C57BL/6J mice. CFU per lung were determined at 48 hpi using BHI agar plates containing X-Gal. All CFU differences between individual infections and corresponding coinfections (A–E, all time points) are statistically significant (P ≤ 0.01; unpaired t test). (B–E) Kinetics of infection for pCD1⁻ mutants (YP6, red), wild-type (YP160, black), and a 1:1 coinfection of pCD1⁻ mutants (blue) and wild-type Y. pestis (green) infected intranasally (B and C) or subcutaneously (D and E). CFU per tissue were determined at various time points of infection for lungs (B), spleen (C and E), and cervical lymph nodes (D). Each bar (A) or point (B–E) represents the mean CFU recovered from five mice. The limit of detection is represented by a dotted line. Data and statistical conclusions are representative of at least two independent experiments. Error bars represent SEM for the bar graph and SD from the mean for line graphs.

These results were somewhat unexpected based on previously published work with closely related Yersinia pseudotuberculosis, from which Y. pestis has rather recently evolved (11). Respiratory coinfection with wild-type Y. pseudotuberculosis is unable to rescue bacterial replication defects of individual Yop mutants (12). All pathogenic Yersinia harbor a nearly identical Ysc T3SS, which directly injects into host cells a number of cytotoxins or effector proteins (Yops) that inhibit bacterial phagocytosis and other innate immune processes (13). Therefore, Y. pestis-specific evolutionary adaptions beyond the Ysc-Yop T3SS have produced a pathogen that is not only substantially more virulent than its immediate ancestor, but also able to compensate for the virulence defects of a coinfecting attenuated strain. We next tested the extent to which fully virulent Y. pestis could compensate for deficiencies of other attenuated isogenic mutants. In these respiratory coinfections, wild-type Y. pestis could also compensate for (or “trans-complement”) the major virulence defects of both Δpla and pCD1⁻ mutants (Fig. 1A), restoring four to six logs of bacterial growth capacity during the first 48 hours postinfection (hpi).

The profound impact of the presence of wild-type Y. pestis on the fate of pCD1⁻ mutants was particularly surprising, because the pCD1 plasmid (commonly called the Yersinia “virulence plasmid”) encodes the entire Ysc T3SS and its virulence-associated Yop effectors. In our intranasal model of infection, pCD1⁻ mutants are severely attenuated (3), unable to proliferate in the lung, and successfully cleared by innate immune mechanisms. Therefore, we examined the impact of wild-type Y. pestis on coinfecting pCD1⁻ mutants in more detail. Although the kinetics of bacterial replication progressed as expected during pulmonary infection with either wild-type bacteria or pCD1⁻ mutants when inoculated individually (3), growth of the pCD1⁻ strain increased substantially when coinoculated at a 1:1 ratio with wild-type Y. pestis. As the infection progressed, this trans-complementation of pCD1⁻ mutant growth continued to increase, reaching six to eight logs at 60 hpi (Fig. 1B). Although we normally do not see pCD1⁻ mutants spreading to the spleen of mice following a pulmonary infection, we were able to detect 10⁴–10⁶ pCD1⁻ bacteria in the spleens of mice coinfected with wild-type Y. pestis at 60 hpi (Fig. 1C).

Trans-Complementation Following Subcutanteous Inoculation.

The lung alveoli play a critical role in respiratory gas exchange, and inflammation-induced edema can lead to severe lung distress. Thus, the lung has many innate immune mechanisms, including the mucociliary escalator and alveolar macrophages, which allow for the rapid clearance of inhaled microorganisms and foreign particles (14). These systems most likely aid in the elimination of Y. pestis pCD1⁻ mutants, which are cleared without producing any appreciable inflammatory response from the host (3). To determine whether unique aspects of the lung's physiology contribute to the ability of Y. pestis to trans-complement avirulent isogenic strains, we also examined the kinetics of bacterial replication in a subcutaneous inoculation model that mimics the route of infection and lymphatic dissemination seen in bubonic plague (Fig. 1D and E). In the draining cervical lymph nodes, pCD1⁻ mutants were able to replicate four to five logs more by 48 hpi when coinfected with wild-type bacteria than when infected individually (Fig. 1D). Wild-type bacteria also facilitated the dissemination of pCD1⁻ mutants to the spleens of animals coinfected by the subcutaneous route (Fig. 1E). These results are consistent with the recent observation of trans-complementation occurring in the spleens (but not the livers) of mice inoculated intravenous with Y. pestis KIM5 (pgm⁻) and an isogenic ΔyopM mutant (15). Although trans-complementation following subcutaneous or intravenous infection are not as pronounced as that seen during pulmonary coinfection, all of these results suggest a dominant immunosuppressive effect of wild-type bacteria and demonstrate that the physiology of different organ systems may determine the extent of trans-complementation.

Localization and Timing of trans-Complementation.

In pneumonic plague, the earliest stages of inflammation appear as distinct foci, and this is also where most of the multiplying bacteria are observed (3). We wondered whether the trans-complementation effect is similarly localized, or whether it reflects immunosuppression of the entire pulmonary compartment. Our first indication came from coinfection experiments in which we varied the inoculum ratios of avirulent to virulent Y. pestis (Fig. 2A). After the first 24 hpi, the numbers of pCD1⁻ mutants never exceeded the numbers of wild-type Y. pestis and were maintained at a constant ratio of 1 pCD1⁻:10 wild-type, regardless of the initial ratio in the inoculum. This tight regulation of proliferation of pCD1⁻ mutants by wild-type Y. pestis suggests that trans-complementation may reflect more of a local control rather than a global effect throughout the lung. In agreement with these observations were the results of pulmonary coinfections with gfp-expressing pCD1⁻ mutants and rfp-expressing wild-type Y. pestis (1:1). Colocalization of the green and red fluorescence revealed replicating pCD1⁻ mutants only in close association with wild-type Y. pestis in inflammatory foci, but not in regions of the lung were wild-type bacteria were not present (Fig. 2B). These data suggest that during the initial stages of infection, an avirulent mutant's proximity to fully virulent Y. pestis determines whether it will proliferate or be eliminated by the lung's innate immune mechanisms.

Fig. 2.

Relative numbers and distribution of virulent and avirulent Y. pestis during pulmonary coinfections. (A) Kinetics of infection for C57BL/6J mice intranasally infected with varying ratios of pCD1⁻ mutants (YP6) and wild-type Y. pestis (YP160) [10:1 (black), 1:1 (blue), and 1:10 (orange) pCD1⁻: wild-type]. CFU per lung were determined at various time points of infection. Each point represents the mean CFU recovered from five mice. Each ratio maintained a 1:10 mutant: wild-type ratio after 24 hpi. There were no statistically significant differences between the CFU of pCD1⁻ mutants between any given ratio at each time point (P > 0.05; unpaired t test). The limit of detection is represented by a dotted line. Data and statistical conclusions are representative of at least two independent experiments. Error bars represent SD from the mean. (B) Mice were coinoculated intranasally with gfp-expressing pCD1⁻ mutants (BGY19) and rfp-expressing wild-type Y. pestis (YP337). At 48 hpi, lungs were fixed and cryosectioned. Fluorescence deconvolution images were generated from 10-μm sections. The image is representative of multiple independent experiments. (Scale bar, 10 μm.)

Trans-Complementation Is Unique to Y. pestis.

The ability of a microorganism to suppress the lung's innate immune system, effectively compensating for mutants missing a variety of virulence factors, is unprecedented among bacterial pathogens. In contrast to the trans-complementation phenomenon we have documented with fully virulent Y. pestis (Fig. S2), wild-type Y. pseudotuberculosis IP2666c was unable to trans-complement isogenic mutants lacking the Yersinia virulence plasmid (designated pYV in Y. pseudotuberculosis) or pCD1⁻ mutants of Y. pestis during pulmonary coinfection (Fig. 3A) (12). Wild-type Y. pseudotuberculosis IP32953, which causes a lung infection with nearly the same bacterial burden as Y. pestis, was also unable to trans-complement an isogenic pYV⁻ strain (Fig. S3A). Similarly, wild-type Klebsiella pneumoniae, which can also cause an acute bacterial pneumonia, failed to significantly trans-complement isogenic strains that have mutations in predominant virulence genes (such as a ΔcpsB mutant, which lacks the polysaccharide capsule) or pCD1⁻ mutants of Y. pestis (Fig. 3B) (16). Conversely, virulent Y. pestis was able to significantly trans-complement ΔcpsB K. pneumoniae and fully trans-complement pYV⁻ Y. pseudotuberculosis IP2666c (Fig. 3C) during pulmonary coinfections. In addition, we have observed that pulmonary infection with wild-type Y. pestis, but not pCD1⁻ mutants, frequently leads to secondary bacterial infections. These bacteria, most likely introduced during the process of intranasal inoculation, can reach levels of 10⁷ bacteria/lung at 60 hpi in wild-type-infected mice (Fig. 3D).

Fig. 3.

Specificity of bacterial trans-complementation during pulmonary infection. (A) Y. pseudotuberculosis, which harbors the same T3SS as Y. pestis, is unable to trans-complement in the lung. Individual intranasal infections (white bars) for pYV⁻ Y. pseudotuberculosis IP2666 mutants (BGY39), pCD1⁻ Y. pestis mutants (BGY14-1), and wild-type Y. pseudotuberculosis IP2666 (BGY30) and matched coinfections (gray bars) with wild-type Y. pseudotuberculosis IP2666 in C57BL/6J mice (all differences between individual and co-infections were not significant, P > 0.05; unpaired t test). (B) K. pneumoniae, another pulmonary pathogen, is unable to trans-complement in the lung. Individual intranasal infections (white bars) for ΔcpsB K. pneumoniae mutants (VK060), pCD1⁻ Y. pestis mutants (BGY14-1) and wild-type K. pneumoniae (KPPR1), and matched coinfections (gray bars) with wild-type K. pneumoniae in C57BL/6J mice (all differences between individual and co-infections were not significant, P > 0.05; unpaired t test). (C) Y. pestis is able to trans-complement avirulent strains of other pathogens. Individual intranasal infections (white bars) for ΔcpsB K. pneumoniae mutants (VK060), pYV⁻ Y. pseudotuberculosis IP2666 mutants (BGY38) and wild-type Y. pestis (YP160), and matched coinfections (gray bars) with wild-type Y. pestis in C57BL/6J mice (all differences between individual and co-infections were significant, P ≤ 0.001; unpaired t test). (D) Wild-type Y. pestis creates a permissive environment that allows nonpathogenic bacteria to proliferate in the lung. Kinetics of infection for pCD1⁻ mutants (red) and wild-type Y. pestis (black) in C57BL/6J mice infected intranasally. CFU per lung were determined at various time points of infection on both BHI (Y. pestis) and BHI-MOX agar plates (non-Y. pestis bacteria). Each bar (A–C) or point (D) represents the mean CFU recovered from five mice. The limit of detection is represented by a dotted line. Data and statistical conclusions are representative of at least two independent experiments. Error bars represent SEM for bar graphs and SD from the mean for line graphs.

Some Yop effector proteins secreted by the plasmid-encoded Ysc T3SS are known to have anti-inflammatory mechanisms, and the Ysc-Yop system is required for all three Yersinia species to cause mammalian infection. Because we observed trans-complementation for Y. pestis but not Y. pseudotuberculosis, even when growing to comparable levels in the lung, we tested whether this could be because of Y. pestis-specific evolutionary changes in the Ysc-Yop T3SS encoded on the virulence plasmid. We exchanged pCD1 in Y. pestis with its corresponding Y. pseudotuberculosis virulence plasmid (pYV), and this Y. pestis pCD1⁻ pYV⁺ strain still trans-complemented a Y. pestis pCD1⁻ mutant (Fig. S3B). These data suggest that although the Ysc-Yop T3SS is necessary for a productive infection and for trans-complementation, subtle evolutionary adaptions to the Ysc-Yop system during the evolution of Y. pestis are not solely responsible for the trans-complementation effect.

Scope of trans-Complementation.

Finally, we wanted to use a broad, unbiased approach to evaluate the scope of trans-complementation by Y. pestis. Negative selection screens are often used to identify and confirm virulence-associated genes in animal models of infection (17). These screens rely on the fundamental assumption that during an infection with a pool of isogenic mutants, it will be possible to select against and subsequently identify single attenuated mutants. We used the microarray-based negative selection technique known as transposon site hybridization (TraSH) (18) to comprehensively characterize trans-complementation during experimental pneumonic plague. The typical expectation of TraSH is that it should reveal in vivo selection against mutants that lack key virulence factors; however, our Y. pestis coinfection results predicted that we may not observe any in vivo selection against individual attenuated mutants in a mixed pool. Remarkably, all known Yersinia virulence genes implicated in pneumonic plague either directly or indirectly showed less than a twofold negative selection in vivo (Table 1). In contrast, negative selection screens have been used to identify and confirm virulence determinants in other lung pathogens, including Pseudomonas aerginosa, Burkholderia pseudomallei, Burkholderia cenocepacia, Streptococcus pneumoniae, and Mycobacterium tuberculosis (19–24).

Table 1.

TraSH screen reveals global trans-complementation during primary pneumonic plague

Gene	Reference	Fold-representation	P value
yopH	(31)	−1.85	0.016
yopD	(13)	−1.33	0.0029
pla	(6)	−1.32	0.076
lcrF	(32)	−1.26	0.030
ail	(33)	−1.88	0.056
psn	(34)	−1.54	0.072

Representatives of known Yersinia virulence genes required for primary pneumonic plague and their representation in the in vivo mutant pool (TraSH probes for the mutant library 48 h postintranasal inoculation in C57BL/6 mice) compared with the in vitro pool (TraSH probes for the mutant library). Negative values indicate an underrepresentation in the output pool.

These observations suggest that Y. pestis is able to create a dominant, localized anti-inflammatory state early during lung infection, and this establishes a unique protective environment that even allows nonpathogenic organisms to prosper. Y. pestis rather recently evolved from a food-borne pathogen (Y. pseudotuberculosis) to become an arthropod-borne pathogen, and this is thought to have strongly favored the coevolution of increased virulence for mammals (25). Poor arthropod vector competence imposed the need to generate very high levels of bacteremia from very few inoculating bacteria to efficiently maintain the natural route of transmission between arthropods and their mammalian hosts. Thus, an early anti-inflammatory state at the site of infection would be biologically advantageous in ensuring that Y. pestis can replicate unfettered until its numbers are sufficient for transmission. When finally recognized by the immune system, the bacteria are at such a high titer that they trigger an overly robust inflammatory response. In the case of a lung infection, the switch from unrestricted bacterial growth to overwhelming inflammation results in rapid decline in pulmonary function and the corresponding fatalities associated with primary pneumonic plague.

Materials and Methods

Reagents, Bacterial Strains, and Culture Conditions.

All chemicals, unless otherwise noted, were obtained from Sigma-Aldrich. Table S1 provides a description of bacterial strains, plasmids and oligonucleotides used in this study. The fully virulent, wild-type Yersinia pestis strain CO92 (YP3) and its plasmid cured derivative CO92 pCD1⁻ (YP6) were obtained from the US Army, Fort Detrick, MD (Table S1). The presence or absence of pCD1, pMT1, pPCP1, and the pgm locus was confirmed by PCR for each strain before use. Y. pestis were routinely grown on brain-heart infusion (BHI) agar (Difco Laboratories) at 26 °C for 2–3 d. Liquid cultures of Y. pestis were grown in BHI broth (Difco Laboratories) at 26 °C for 6–8 h before being diluted to an OD₆₂₀ of 0.05–0.1 in 10 mL of BHI supplemented with 2.5 mM CaCl₂ in a 125-mL Erlenmeyer flask and grown overnight at 37 °C or 26 °C on a shaker set at 250 rpm. Y. pseudotuberculosis IP2666c and derivatives were obtained from the laboratory of Joan Mecsas (Tufts University, Boston, MA). Y. pseudotuberculosis IP32953 and K. pneumoniae subspecies pneumoniae and derivatives were obtained from the laboratory of Virginia Miller (University of North Carolina at Chapel Hill). Y. pseudotuberculosis strains were routinely grown at 26 °C on LB agar for 2 d or in LB broth for 12–14 h. K. pneumoniae strains were grown at 37 °C for 10–12 h on LB agar or in LB broth. For animal infections, Y. pseudotuberculosis and K. pneumoniae were grown in BHI broth as described above for Y. pestis. Media were supplemented with ampicillin (100 μg⋅mL⁻¹), kanamycin (50 μg⋅mL⁻¹), polymyxin B (25 μg⋅mL⁻¹), X-Gal (40 μg⋅mL⁻¹), or magnesium oxalate (MOX; 20 mM MgCl₂ and 20 mM Na₂C₂O₄) as required. Chelation of calcium ions by MOX causes a “low calcium response” restriction of 37 °C growth of Y. pestis harboring pCD1.

Construction of Y. pestis Deletion Strains.

Y. pestis deletion strains (YP373 and BGY14-1) were constructed using a modified lambda red recombination system described previously by Lathem et al. (6). Briefly, 500 bp upstream and 500 bp downstream of each gene were amplified by PCR using the oligonucleotides yopH 3′+500 and P1yopH 3′3 (yopH upstream region), yopH 5′-500 and P4yopH 5′1406 (yopH downstream region), YplacZUP-F and YplacZUP-R (lacZ upstream region), and YplacZDN-R and YplacZDN-F (lacZ downstream region). The respective PCR products were gel-purified and combined in splicing by overhang extension (SOE)-PCR reactions with a kan^R cassette flanked by FRT sites, which had previously been applied from the plasmid pKD13 (26). Oligonucleotides yopH 3′+500, yopH 5′-500, YplacZUP-F and YplacZDN-R were used in their respective SOE-PCR reactions. The resulting PCR products (yopH-FRT-Kan^R-FRT-yopH and lacZ-FRT-Kan^R-FRT-lacZ) were gel-purified and transformed into YPI54 or BGY0, Y. pestis strains carrying pWL204, which had been grown at 26 °C in the presence of 10 mM arabinose to induce the lambda red recombinase genes. Putative recombinants were selected on BHI plates containing kanamycin. Recombinants were passaged on BHI plates containing 5% sucrose to cure pWL204. The kan^R cassette in the ΔyopH mutant strain was resolved by transforming the strain with pSkippy, a tet^S derivative of pFLP3 (27) carrying the FLP recombinase and sacB genes, and growing the strain overnight at 26 °C in the presence of 1 mM isfopropyl-β-d-thiogalactopyranoside. The overnight cultures were diluted and plated on BHI plates containing 5% sucrose to cure pSkippy. Kan^S and amp^S recombinants were selected and in-frame deletions of the target gene were verified by PCR. For each strain, the presence of all of the major virulence loci (pCD1⁺, pMT1⁺, pPCP1⁺ and pgm⁺) was confirmed by PCR.

Construction of Y. pseudotuberculosis IP2666c Strains.

Y. pseudotuberculosis IP2666c pIB1⁻ (BGY39) was constructed by passaging the Y. pseudotuberculosis IP2666c pIB1⁺ (LL-89) on BHI-MOX agar plates and selecting large colony variants, indicating the loss of the virulence plasmid pIB1. The loss of pIB1 was confirmed using PCR. Y. pseudotuberculosis pYV⁻ lacZ::kan^R (BGY38) was constructed via lambda red recombination as described above, using the same primers to amplify the regions adjacent to lacZ.

Animals.

All animal studies were approved by the University of North Carolina in Chapel Hill Office of Animal Care and Use protocol #09–057.0-A or the Washington University Animal Studies Committee protocol #20060154. Pathogen-free, 6- to 8-wk-old C57BL/6J mice were obtained from The Jackson Laboratory. Mice were housed in a Techniplast Isocage containment system during all experiments. Mice were provided with food and water ad libitum and maintained on a 12-h light/12-h dark cycle at 25 °C and 15% humidity. Bacteria were grown in BHI broth as described above, and the optical density (OD₆₂₀) of the culture was used to estimate bacterial concentrations. Actual bacterial concentrations were determined by plating serial dilutions of the inoculum on BHI or LB agar plates. The bacteria were then washed in endotoxin-free PBS and maintained at 37 °C (intranasal infection) or 26 °C (subcutaneous infection) before inoculation, unless otherwise indicated. Mice were anesthetized with a ketamine/xylazine solution (50 mg⋅kg⁻¹/10 mg⋅kg⁻¹) and inoculated with 20 μL or 100 μL of bacteria in PBS via an intranasal or subcutaneous route, respectively. Animals were monitored twice daily and killed with an overdose of sodium pentobarbital (180 mg⋅kg⁻¹) at designated timepoints, or when the animal was clearly moribund.

Animal Infections.

For both the individual infection and coinfection experiments, groups of four to five mice were inoculated intranasally with 1 × 10⁴ CFU of Y. pestis, Y. pseudotuberculosis, or K. pneumoniae for individual and coinfections with equal ratios (1 pCD1⁻:1 wild-type). In experiments with varying ratios (1 pCD1⁻:10 wild-type, 1 pCD1⁻:1 wild-type and 10 pCD1⁻:1 wild-type), we inoculated with1 × 10⁵ CFU of Y. pestis. Groups of four to five mice were inoculated subcutaneously in the front neck with 1,000 CFU of Y. pestis for both individual and coinfections. For both Y. pestis and Y. pseudotuberculosis, lacZ mutants (10, 12) were designated as our “wild-type” strain, and blue/white screening was used to differentiate between coinoculated strains. Bacterial organ burdens were determined by plating bacteria on BHI or LB agar containing X-Gal for experiments involving multiple Y. pestis or Y. pseudotuberculosis strains. For experiments involving K. pneumoniae, organ burdens were determined using LB agar with kanamycin (selection mutant) or without kanamycin. For mixed species experiments [Y. pseudotuberculosis or K. pneumoniae with Y. pestis pCD1⁻ lacZ::kan^R (BGY14-1)], organ burdens were determined using BHI agar with kanamycin (Y. pestis pCD1⁻ mutant selection) or without kanamycin. At the indicated times postinoculation, mice were killed, lungs and spleen (after intranasal inoculation) or lymph nodes, spleen, and lungs (after subcutaneous inoculation) were surgically removed, weighed, and homogenized in 0.5 mL PBS. Serial dilutions for each organ were plated on BHI or LB agar as indicated above. Bacterial burdens are reported as CFU/tissue.

Histopathology.

Groups of three to five mice were inoculated intranasally with 1 × 10⁴ CFU of a 1:1 ratio of Y. pestis [1 gfp-expressing pCD1⁻ (BGY23):1 rfp-expressing wild-type (YP337)]. Mice were killed at 48 hpi with an overdose of sodium pentobarbital, and their lungs were inflated with 10% neutral buffered formalin via tracheal cannulation. The lungs were then removed from the animal and fixed for 96 h in 10% neutral buffered formalin. All samples were negative for bacterial growth before being removed from the BSL-3 facility. The lungs were washed twice in PBS for 1 h and then in PBS-30% sucrose for 24 h. Frozen sections were generated by freezing the lungs in optimal cutting temperature compound on dry ice for 20 min and then at −80 °C overnight. Ten-micrometer sections were mounted with Pro-Long Gold (Invitrogen) and viewed under an Olympus BX60 fluorescence microscope. iVision software v. 4.0.0 (BioVision Technologies) was used for z-stack image collection and deconvolution.

Construction of Transposon Mutant Library.

The transposon mutant library was generated using the Himar1-based transposon system encoded on pPP47, which was constructed in multiple steps using the phagemid φMycoMarT7 (28) and pTnMod-RKm (29). Briefly, pTnMod-RKm was digested with BglII, and blunt ends were generated using Klenow treatment. The resulting fragments were then digested with EcoRI, and the fragment corresponding to the RP4 origin of transfer was ligated to the Himar1 transposon, which was previously amplified from φMycoMarT7 using the oligonucleotides TraSHHimarIR5′(PmeI/AscI) and TraSHIR3′(MfeI), and transformed into Escherichia coli S17. The Himar1 transposon contains an R6K origin of replication, a Kan^R cassette and flanking T7 promotors oriented so as to promote transcription into the chromosomal DNA. The Tn5 promotor region (pTnMod-RKm) and the Himar1 transposase (φMycoMarT7) were independently amplified using the oligonucleotides TraSH-TnModPromII and TraSH Tase-PromII, and TraSHTase-PromIIGC and TraSH5′Tase(AscI), respectively, and then joined together by SOE-PCR. The resulting PCR products were cloned into the AscI site adjacent to the Himar1 transposon to create pPP47. E. coli S17 carrying pPP47 were mated with wild-type Y. pestis and plated on BHI agar containing kanamycin for Y. pestis selection and polymyxin B for E. coli counter selection. Southern blot analysis of 20 randomly selected colonies indicated that each mutant contained a single transposon insertion. There were ∼2.5 × 10⁵ mutants in the transposon mutant library used in the subsequent experiments.

Transposon Site Hybridization.

The transposon mutant library was used in animal experiments to identify genes that are required for infection during primary pneumonic plague. The protocols used in these experiments have been described previously (18, 28). Briefly, 10 mice were inoculated intranasally with 2 × 10⁵ CFU of the transposon mutant library. The in vitro pool was generated by replating the library on BHI agar. At 48 hpi, the mice were killed, and a subset of the surviving bacteria (5 × 10⁵ CFU) was recovered from lung homogenates by plating on BHI agar plates containing kanamycin. Genomic DNA was isolated from each pool, and modified TraSH probes were generated as previously described (28). Briefly, genomic DNA was partially digested with HinPI and MspI. DNA fragments (500–2,000 bp) were gel-purified and ligated to adaptors generated with the oligonucleotides TraSH-Probe-Lig1 and TraSH-Probe-Lig2. Ligated DNA was used as a template in two separate PCR reactions containing the oligonucleotide TraSH-Probe-PCRadaptor and either TraSH-Probe-PCRtrans1 or TraSH-Probe-PCRtrans1. Gel-purified PCR products (250–500 bp) were used template for in vitro T7 transcription reactions (Ambion). Template DNA was digested with TurboDNase (Ambion), and the RNA was purified and used to generate ds-cDNA (Invitrogen) with the primers TraSH-Probe-PCRadaptor and TraSH-cDNA-2^nd-strand. The in vitro and in vivo probes were then labeled with Cy3 or Cy5 and cohybridized to the array as described previously (3). Probes were generated from each pool three times and analyzed on duplicate microarrays for a total of six microarray hybridizations per experiment. Two independent experiments were performed, yielding consistent results. The arrays were scanned following hybridization and analyzed using Partek Genomics Suite (Partek Inc). Data for each experiment were analyzed using multiple testing correction after three-way ANOVA and reported as the fold-representation in the in vivo pool compared with the in vitro pool, where negative values indicate a relative underrepresentation in the in vivo pool. This approach was technically validated by defining the genes required for Y. pestis growth on BHI agar plates, similar to validation performed previously for other organisms (30).

Statistical Analyses.

An unpaired Student t test was used to determine the statistical significance (P ≤ 0.05) between individual and coinfection experiments. If the variance between datasets was statistically significant, the data were transformed (log-based transformation) to achieve normality between datasets before calculating their significance. All statistical analyses were calculated using Prism v4.0c software (GraphPad Software).