New Insights into Autoinducer-2 Signaling as a Virulence Regulator in a Mouse Model of Pneumonic Plague.

The Enterobacteriaceae family members, including the infamous Yersinia pestis, the causative agent of plague, have a highly conserved interbacterial signaling system that is mediated by the autoinducer-2 (AI-2) quorum-sensing molecule. The AI-2 system is implicated in regulating various bacterial virulence genes in diverse environmental niches. Deletion of the gene encoding the synthetic enzyme for the AI-2 substrate, luxS, leads to either no significant change or, paradoxically, an increase in in vivo bacterial virulence. We showed that deletion of the rbsA and lsrA genes, components of ABC transport systems that interact with AI-2, synergistically disrupted AI-2 signaling patterns and resulted in a more-than-50-fold decrease in Y. pestis strain CO92 virulence in a stringent pneumonic plague mouse model. Deletion of luxS or lsrK (encoding AI-2 kinase) from the ΔrbsA ΔlsrA background strain or complementation of the ΔrbsA ΔlsrA mutant with the corresponding gene(s) reverted the virulence phenotype to that of the wild-type Y. pestis CO92. Furthermore, the administration of synthetic AI-2 in mice infected with the ΔrbsA ΔlsrA ΔluxS mutant strain attenuated this triple mutant to a virulence phenotype similar to that of the ΔrbsA ΔlsrA strain in a pneumonic plague model. Conversely, the administration of AI-2 to mice infected with the ΔrbsA ΔlsrA ΔluxS ΔlsrK mutant did not rescue animals from lethality, indicating the importance of the AI-2-LsrK axis in regulating bacterial virulence. By performing high-throughput RNA sequencing, the potential role of some AI-2-signaling-regulated genes that modulated bacterial virulence was determined. We anticipate that the characterization of AI-2 signaling in Y. pestis will lead to reexamination of AI-2 systems in other pathogens and that AI-2 signaling may represent a broad-spectrum therapeutic target to combat antibiotic-resistant bacteria, which represent a global crisis of the 21st century. IMPORTANCEYersinia pestis is the bacterial agent that causes the highly fatal disease plague. The organism represents a significant concern because of its potential use as a bioterror agent, beyond the several thousand naturally occurring human infection cases occurring globally each year. While there has been development of effective antibiotics, the narrow therapeutic window and challenges posed by the existence of antibiotic-resistant strains represent serious concerns. We sought to identify novel virulence factors that could potentially be incorporated into an attenuated vaccine platform or be targeted by novel therapeutics. We show here that a highly conserved quorum-sensing system, autoinducer-2, significantly affected the virulence of Y. pestis in a mouse model of pneumonic plague. We also identified steps in autoinducer-2 signaling which had confounded previous studies and demonstrated the potential for intervention in the virulence mechanism(s) of autoinducer-2. Our findings may have an impact on bacterial pathogenesis research in many other organisms and could result in identifying potential broad-spectrum therapeutic targets to combat antibiotic-resistant bacteria, which represent a global crisis of the 21st century.

INTRODUCTION

Autoinducer-2 (AI-2), a quorum-sensing (QS) molecule found widely among Gram-positive and -negative bacteria, is associated with a diverse array of virulence mechanisms, ranging from secretion systems to biofilm formation in in vitro culture assays (1–7). Despite the linking of virulence mechanisms to AI-2 signaling, evidence of biological significance for these signaling pathways is limited in in vivo models (4, 7–9). Generally, the AI-2 signaling is characterized in a given organism by deleting the gene encoding the primary synthetic enzyme for the AI-2 substrate, LuxS, and observing changes in bacterial virulence phenotypes (10). During the course of our investigation into novel virulence factors of Yersinia pestis, the causative agent of plague, we reported a dramatic increase in attenuation of the Δlpp ΔmsbB ΔrbsA combinatorial deletion mutant in a stringent pneumonic plague mouse model (11). Our earlier studies showed that deletions of lpp, the gene encoding Braun lipoprotein (Lpp), and msbB, a gene encoding a lipopolysaccharide (LPS)-modifying acyltransferase (MsbB), attenuated a highly virulent Y. pestis strain, CO92 (12–14). While Lpp activates Toll-like receptor 2 (TLR-2) signaling, MsbB adds lauric acid to the lipid A moiety of LPS to modulate TLR-4 signaling (12). The additional deletion of rbsA (identified during our genome-wide, transposon-based, signature-tagged mutagenesis of Y. pestis CO92 [11]), encoding the ATP binding protein ribose ATP binding cassette (ABC) transporter, led to a further attenuation of the Δlpp ΔmsbB mutant that was in excess of 10-fold (11). Investigation into the mechanism of the attenuation due to the deletion of rbsA within the rbsBAC operon showed that RbsA was necessary for efficient bacterial growth in a minimal medium limited to a ribose carbon source (11). While RbsA has ATPase activity, its coupling with RbsC, a bacterial membrane-associated protein, actively transports ribose that has been shuttled through the periplasm of the organism by high-affinity association with RbsB (15, 16).

In addition to the role in ribose utilization, orthologs of ribose transport proteins, such as RbsB in Aggregatibacter actinomycetemcomitans, efficiently interacted with AI-2 under physiologically relevant conditions (2, 9). The ribose transporter (Rbs), as well as the Lsr (LuxS-regulated) ABC transporter, are responsible for the uptake of the AI-2 QS signaling molecule into bacterial cells in many pathogenic bacteria which do not possess the dedicated two-component circuit of Vibrio harveyi (17). V. harveyi produces three autoinducers: AI-1 (3-hydroxybutanoyl homoserine lactone), CAI-1 [(S)-3-hydroxytridecan-4-one], and AI-2 [(2S,4S)-2-methyl-2,3,3,4-tetrahydroxytetrahydrofuranborate] (18), which are detected extracellularly by their cognate transmembrane receptors, namely, LuxN, CqsS, and LuxPQ, respectively (18). Signals through the autoinducer-sensing pathways are then transduced through shared components LuxU and LuxO and five small regulatory RNAs (sRNAs) to the master quorum-sensing regulator LuxR in V. harveyi (19, 20). An earlier study of AI-2 in an attenuated Y. pestis strain, KIM 1001 (with a deletion of the pigmentation locus [pgm] required for iron uptake), revealed significant expression changes in large sets of genes, as well as diminished oxidative damage resistance, when luxS was deleted from the Δpgm mutant (7). The luxS gene encodes the AI-2 synthetic enzyme, while the lsrK gene encodes a kinase which phosphorylates AI-2, and the sequestered phospho-AI-2 then binds to the LsrR repressor to activate transcription of the lsr operon (21). However, deletion of the luxS gene from a fully virulent KIM5 strain of Y. pestis did not alter the 50% lethal dose (LD50) compared to that of the wild-type (WT) bacterium in a mouse model of bubonic plague (22). In this study, we demonstrated for the first time that the disruption of AI-2 transport from the extracellular milieu into Y. pestis CO92 due to the deletion of the rbsA and lsrA genes resulted in a significant reduction of virulence of the mutant in a mouse model of pneumonic plague. Furthermore, the deletion of the luxS or lsrK gene compromised the attenuated phenotype of the ΔrbsA ΔlsrA mutant, thus providing new insights into AI-2 signaling.

RESULTS

Deletion of rbsA and lsrA genes from Y. pestis CO92 disrupts autoinducer-2 signaling.

The initial finding we reported, that the deletion of the rbsA gene synergistically attenuated Y. pestis CO92 in association with deletions of lpp and msbB genes in a mouse model of pneumonic plague, led us to investigate mechanisms of attenuation beyond the impairment of ribose transport and utilization (11). Since orthologs of the Rbs operon are also associated with AI-2 transport, we examined the effect of combinatorial deletion of lpp, msbB, and rbsA on the levels of AI-2 in the culture supernatants of mutants versus the level in the supernatant of WT CO92. At temperatures of both 28°C (flea) and 37°C (human body), representing two lifestyles of Y. pestis (23), there were major aberrations in the patterns of AI-2 in the highly attenuated mutants (the Δlpp ΔmsbB and Δlpp ΔmsbB ΔrbsA mutants) compared to its occurrence in WT CO92 that were independent of the bacterial growth rates (data not shown). Interestingly, deletion of rbsA from the Δlpp ΔmsbB mutant had a potentiating effect on disrupting AI-2 patterns. This initial finding suggested that changes in AI-2 were potentiated with deletion of the rbsA gene but were inadequate to indicate a causal relationship, leading us to study the AI-2 signaling system in greater detail.

Since the deletion of the lpp and msbB genes from Y. pestis could possibly affect the topology of the bacterial membrane and/or the expression of the stress response genes (23), we evaluated changes in the AI-2 levels of mutants with a deletion in the canonical AI-2 transport system both singly (lsrA) and in combination with rbsA. The lsrA gene, which encodes the ATPase component of the Lsr ABC transporter complex, was chosen because of the similarity of its function to that of RbsA, which is also an ATPase for the RbsBAC transporter complex. When the AI-2 levels were measured in serially diluted culture supernatants using the V. harveyi reporter strain BB170, the single deletion of rbsA resulted in no discernible effect on the AI-2 pattern when compared to that of WT CO92 at both temperatures, 28°C and 37°C (Fig. 1A and B). As previously reported (7), the lsrA deletion resulted in increases of AI-2 during stationary phase (after 22 h); however, when combined with the rbsA deletion, the ΔrbsA ΔlsrA mutant strain exhibited greater-than-two- to threefold increases in the levels of extracellular AI-2 during the mid- to late log phase of growth, as well as potentiation of the stationary-phase changes in the AI-2 levels observed with the ΔlsrA single mutant (Fig. 1A and B). The increases in extracellular AI-2 were observed both at 28°C and 37°C, with similar magnitudes of AI-2 increase.

FIG 1 

AI-2 levels in cell-free supernatants of WT Y. pestis CO92 and its various mutant strains. (A and B) Concentrations of AI-2 in culture supernatants during growth at 28°C (A) or 37°C (B) are shown, with absorbance of culture at 600 nm in insets. (C) AI-2 levels in cell-free supernatants following dosing of cultures with synthetic exogenous AI-2. Data are from three independent cultures per strain. Arithmetic means ± standard deviations from three independent experiments are shown. Statistical analysis was performed using pairwise t tests with multiple comparison correction.

The above-described changes in the extracellular levels of AI-2 could be the result of two competing processes: increased synthesis of AI-2 through the LuxS enzyme or decreased uptake of AI-2 from the extracellular milieu or both. To discriminate between these two possibilities, the synthesis of AI-2 was eliminated by the deletion of the luxS gene. The ΔluxS mutant was able to efficiently take up synthetic AI-2 from the culture medium during in vitro growth (Fig. 1C); however, the ΔrbsA ΔlsrA ΔluxS mutant was impaired in its ability to transport synthetic AI-2, with statistically significant delayed uptake and altered kinetics that led to a slow linear decrease in the concentration of AI-2 in the culture supernatant. The ΔluxS mutant with intact rbs and lsr operons exhibited a sigmoidal depletion of AI-2 that was initiated approximately 1 h prior to the ΔrbsA ΔlsrA ΔluxS mutant’s lower rate of AI-2 uptake (Fig. 1C).

Changes in AI-2 signaling correlate to in vitro and in vivo attenuation of Y. pestis CO92.

Following the confirmation of AI-2 aberrations, we determined whether there was any correlation between changes in AI-2 signaling and in vitro virulence as measured by intracellular survival (ICS) of the mutants in RAW 264.7 murine macrophages compared to the survival of WT CO92 (Fig. 2A). The ability to survive and replicate within macrophages and the recruitment of early immune effector cells during infection contribute to pathogenicity in in vivo models and, as such, are important measures of Y. pestis virulence (24, 25). We found that the single deletion of either of the transport protein-encoding genes (rbsA or lsrA) had a minimal effect on ICS, but when they were combined, the ΔrbsA ΔlsrA mutant was significantly less resistant to the macrophage intracellular environment (Fig. 2A). All of the mutants tested exhibited levels of phagocytosis similar to that of WT CO92 (Fig. 2B).

FIG 2 

Intracellular survival of WT Y. pestis CO92 and its various mutant strains in macrophages. (A) The intracellular survival of WT Y. pestis and its various mutants in a RAW 264.7 murine macrophage cell line was evaluated at 0 and 4 h postinfection (hpi). Luminescent reporter strains from each background were utilized to evaluate real-time reporting of bacterial survival in macrophages. (B) Phagocytosis of bacteria was determined by comparing luminescence of infectious dose after culture to luminescence at 0 h. Data are from three independent experiments per strain. Arithmetic means ± standard deviations are shown. Statistical analysis was performed using one-way analysis of variance (ANOVA) with Tukey post hoc correction.

The data described thus far have been consistent with previous studies of AI-2 signaling in various pathogenic bacteria, showing changes in bacterial virulence related to in vitro assays. However, when we challenged mice in a pneumonic plague model to determine whether these changes in AI-2 signaling correlated with an alteration in in vivo virulence, our findings diverged from the previous literature (Fig. 3) (4, 8). The single gene deletion of either rbsA or lsrA showed only modest decreases in virulence in a mouse model; the survival of mice challenged with the ΔrbsA mutant reached statistical significance at 30% survival when a challenge dose equivalent to 9 LD50 of WT CO92 was used, while the survival of mice challenged with the ΔlsrA mutant was not significantly different (20% survival) from that of the WT CO92-challenged group of mice. The ΔrbsA ΔlsrA combinatorial deletion resulted in a significant decrease in the virulence of the mutant, with 80 to 100% of mice surviving a challenge dose of 8- to 50-LD50 equivalent of WT CO92 (Fig. 3A). These data for the ΔrbsA ΔlsrA mutant were comparable to our published results for virulence attenuation of the Δlpp ΔmsbB ΔrbsA strain, with 100% survival at a 50-LD50 equivalent of WT CO92 (11); the Δlpp ΔmsbB ΔrbsA strain also exhibited changes in AI-2 levels similar to those seen for the ΔrbsA ΔlsrA mutant (data not shown). The attenuated phenotype of the ΔrbsA ΔlsrA mutant could be complemented through a site-specific, single-copy, mini-Tn7 transposon insertion of the native gene, along with the promoter, of either rbsA or lsrA (Fig. 3B). The significant decreases in virulence of the above-described mutants (the ΔrbsA ΔlsrA and Δlpp ΔmsbB ΔrbsA mutants) (11) that we observed in animals was unexpected, given the extensive literature indicating a negligible role for AI-2 in regulating in vivo virulence (4, 7, 8, 10, 26). As such, our results merited a more thorough investigation, and we decided to determine the additional role that luxS plays in virulence.

FIG 3 

In vivo virulence of various Y. pestis mutants. (A and B) Survival of female Swiss-Webster mice (n = 5 to 10) in a pneumonic plague model after challenge with the stated dose (equivalent to WT LD50, where 1 LD50 is 500 CFU) of the rbsA, lsrA, or ΔrbsA ΔlsrA deletion mutant (A) or the ΔrbsA ΔlsrA double deletion strain complemented with the Tn7 transposon (B). Data are representative of three independent experiments. Statistical analysis was performed using Kaplan-Meier survival curve analysis.

Suppression of ΔrbsA ΔlsrA phenotype of decreased bacterial virulence in vivo by luxS deletion in Y. pestis CO92.

To characterize the effect luxS deletion had on the AI-2 signaling pathway, we constructed single and combinatorial deletions of luxS in the background strains of WT CO92 and the ΔrbsA ΔlsrA mutant, in addition to developing various complemented and overexpressing strains. These strains were then characterized by determining the AI-2 levels in the culture supernatants over a period of bacterial growth. As expected, the deletion of luxS led to abrogation of AI-2 at all time points of bacterial growth tested (data not shown), which was replicated by the ΔrbsA ΔlsrA ΔluxS mutant (Fig. 4A). Interestingly, we found that the copy number of the luxS gene was critical in the context of AI-2 uptake during complementation of the ΔrbsA ΔlsrA ΔluxS mutant. trans complementation via a low-copy-number plasmid vector, pBR322, resulted in an early accumulation of AI-2 in the culture supernatant, with altered kinetics of depletion compared to that in the background ΔrbsA ΔlsrA mutant strain (Fig. 4A). Comparable AI-2 levels in the supernatants of the ΔrbsA ΔlsrA ΔluxS mutant and the ΔrbsA ΔlsrA background strain were only observed when complementation was achieved through cis complementation of a single copy of the luxS gene utilizing a Tn7-based transposon system (Fig. 4A).

FIG 4 

In vivo virulence and AI-2 levels in cell-free supernatants of ΔluxS and ΔlsrK mutants. (A and C) Concentrations of AI-2 in culture supernatants over a 30-h period of growth at 28°C. Error bars represent standard deviations from three independent experiments. (B and D) Survival of female Swiss-Webster mice (n = 5 to 10) in a pneumonic plague model after challenge with 10-LD50 equivalent of WT CO92, where 1 LD50 is 500 CFU. Data are representative of three independent experiments. Statistical analysis for animal studies was performed using Kaplan-Meier survival curve analysis.

We then evaluated the luxS deletion strains for their ability to cause disease in a pneumonic plague mouse model. We noted a significant trend not reported before in the literature (Fig. 4B). While the ΔluxS mutant was as virulent as the WT CO92, the deletion of luxS resulted in suppression of the attenuated phenotype of the ΔrbsA ΔlsrA background strain, thus attaining full virulence similar to that of WT CO92 (Fig. 4B). Interestingly, when bacterial CFU were evaluated in the lungs of mice at 24, 48, and 72 h and 7 days postinfection, it was noted that both the WT CO92 and the ΔluxS strain exhibited rapid growth up to 1010 CFU by 72 h postinfection (see Fig. S1A in the supplemental material). However, in the lungs of mice infected with the ΔrbsA ΔlsrA strain, bacterial counts persisted, albeit at much lower numbers and in the range of 102 to 104 CFU, out to 72 h postinfection before being cleared on day 7 postinfection (see Fig. S1A).

Lung colonization and intracellular survival of WT Y. pestis CO92 and its ΔluxS and ΔrbsA ΔlsrA deletion mutants in mice (n = 4). (A) Bacterial lung counts were obtained from whole-lung homogenates harvested at 24, 48, or 72 h or 7 days postinfection with a 10-LD50 dose by the intranasal (i.n.) route in female Swiss-Webster mice. The horizontal line represents average counts from four animals. (B) Intracellular survival of ΔluxS, ΔrbsA ΔlsrA, or ΔrbsA ΔlsrA ΔluxS deletion mutant compared to that of WT CO92. The intracellular survival of WT CO92 and its various mutants in a RAW 264.7 murine macrophage cell line was evaluated at 0 and 4 h postinfection. Luminescent reporter strains from each background were utilized to evaluate real-time reporting of bacterial survival in macrophages. Statistical analysis was performed using one-way ANOVA with Tukey post hoc correction, and three independent experiments were performed. ***, P < 0.001; *, P < 0.01; hpi, hours postinfection; dpi, days postinfection. Download Figure S1, TIF file, 0.3 MB.

We also found that the copy number of luxS modified the observed phenotype of the ΔrbsA ΔlsrA ΔluxS mutant. While the overexpression of luxS via the trans complementation strategy (with pBR322) resulted in a virulent phenotype (Fig. 4B), cis complementation of the ΔrbsA ΔlsrA ΔluxS mutant with luxS using the Tn7-based system led to an avirulent phenotype (80% survival) in a mouse model of pneumonic plague, possibly suggesting a dose-dependent effect of the luxS gene on Y. pestis virulence (Fig. 4B). Overall, our data obtained in the context of the ΔrbsA ΔlsrA mutant indicated that both the loss of luxS, as in the ΔrbsA ΔlsrA ΔluxS mutant, and the overexpression of luxS, as in the ΔrbsA ΔlsrA(pBR-luxS) mutant, led to virulent phenotypes in vivo (Fig. 4B). When we examined the ICS of these strains, we found a similar trend, where a single luxS deletion induced a phenotype similar to that of WT CO92 and suppression of attenuating characteristics in terms of ICS for the ΔrbsA ΔlsrA ΔluxS strain (see Fig. S1B in the supplemental material).

The paradigm of equating LuxS with AI-2 function has been questioned in the past, particularly due to multiple roles of LuxS beyond AI-2 substrate production (10). To verify whether the phenotype observed in luxS mutants was derived from its effects on AI-2, we generated a series of mutants that carried a deletion in the gene coding for the AI-2 kinase, lsrK. Prior literature has demonstrated that the deletion of lsrK results in an Escherichia coli strain that is both insensitive to AI-2 and unable to sequester AI-2, leading to a high concentration of it in the extracellular milieu that does not decrease over time (21, 27). We confirmed that this phenotype occurs in Y. pestis with a deletion of lsrK; interestingly, in the ΔlsrK and ΔrbsA ΔlsrA ΔlsrK mutants, the levels of AI-2 rise to concentrations similar to those seen for the ΔrbsA ΔlsrA mutant but then remain at these elevated levels through the stationary phase (Fig. 4C). As expected, the ΔrbsA ΔlsrA ΔluxS ΔlsrK quadruple mutant did not synthesize any AI-2. When lsrK mutants (ΔlsrK and ΔrbsA ΔlsrA ΔlsrK mutants) were tested for virulence in a pneumonic plague mouse model, they were universally lethal (Fig. 4D). In other words, the attenuated-virulence phenotype of the ΔrbsA ΔlsrA mutant was suppressed with the deletion of the lsrK gene, similar to the result for the luxS gene deletion mutant (Fig. 4B and D).

Thus far, we have shown a correlation between disruptions in AI-2 signaling due to aberrant transport of AI-2 within the bacterial cell and the attenuation of phenotypes in both in vitro and in vivo models of plague. Furthermore, modulation of AI-2 signaling due to the deletion of either the luxS gene or the downstream lsrK gene reverts the attenuated phenotype of the ΔrbsA ΔlsrA mutant to a phenotype similar to that of the WT bacterium.

Transcriptomic profiles of Y. pestis CO92 strains in which AI-2 is perturbed.

To identify potential mechanism(s) of attenuation and further link the observed changes in AI-2 levels to the attenuated phenotypes of the mutants, we subjected each of the major mutant strains to high-throughput RNA sequencing (RNA-seq) analysis. RNA was isolated from the bacterial strains at peak AI-2 levels during the mid- to late exponential phase of growth, when the most significant aberrations in AI-2 signaling were observed (Fig. 1A and B). A heat map of the top 100 most-variable genes showed similar expression patterns within each strain and common expression patterns shared between strains in which AI-2 was perturbed, as well as isolated groupings unique to the attenuated ΔrbsA ΔlsrA strain (Fig. 5A). Distance mapping of the strains revealed a hierarchical grouping of the samples within their strains, exhibiting low variance between samples, as well as commonalities between the ΔrbsA ΔlsrA ΔluxS and ΔluxS strains (Fig. 5B and C).

FIG 5 

Analysis of AI-2 mutant transcriptomes. (A) A heat map was constructed against mean counts of the top 100 most variable genes across all samples, including WT CO92 (WT), ΔrbsA ΔlsrA (RL), ΔluxS, and ΔrbsA ΔlsrA ΔluxS (RLS) mutants. (B and C) Distance maps were created using Poisson (B) or Euclidean (C) distance against the transcriptome of each mutant examined. (D) Relative transcript levels from the RNA-seq data set were plotted for selected genes, with the level of each gene in each strain being compared to its level in WT CO92. (E) Relative expression levels of selected genes from qRT-PCR data were plotted for the indicated strains after growth with exogenous AI-2 added at the indicated concentrations. Standard deviations of the results from 3 independent experiments are shown.

Of the approximately 4,000 genes in the Y. pestis genome, 219 were differentially expressed between the ΔrbsA ΔlsrA strain and WT CO92 by more than twofold up or down at a significance level of Padj of <0.05, with P adjusted for multiple comparisons as defined by DESeq2, 119 were differentially expressed between ΔluxS and WT CO92, 46 between the ΔluxS and ΔrbsA ΔlsrA ΔluxS strains, and 78 between the ΔluxS and ΔrbsA ΔlsrA strain (see Tables S1 and S2 in the supplemental material). From these data, we identified patterns of gene expression unique to the attenuation of the ΔrbsA ΔlsrA strain, changes that suppressed attenuation in the ΔrbsA ΔlsrA ΔluxS strain, and lastly, expression patterns common to both the ΔrbsA ΔlsrA and the ΔluxS mutant.

Significantly differently expressed genes in the ΔrbsA ΔlsrA strain versus WT CO92 and in the ΔluxS strain versus WT CO92. Download Table S1, PDF file, 1 MB.

Significantly differently expressed genes in the ΔluxS strain versus the ΔrbsA ΔlsrA ΔluxS strain and in the ΔluxS strain versus the ΔrbsA ΔlsrA strain. Download Table S2, PDF file, 0.8 MB.

Attenuating expression was identified by sorting for significant changes in the ΔrbsA ΔlsrA strain versus WT CO92 and excluding any significant changes common to the ΔluxS or ΔrbsA ΔlsrA ΔluxS strain and WT CO92 (see Table S3 in the supplemental material). Attenuating changes in expression (mean fold changes in expression are shown in parentheses below) included 249 genes with a Padj of <0.1, comprising genes encoding ABC transporter families specific to arabinose (araCFGH) (−1.8 to −5.9) and galactose (mglABC) (−1.6 to −1.8), chaperone-encoding genes dnaJK (−2.4 to −2.5), ibpAB (−2.6), and htpG (−3.2), oxidative phosphorylation gene family atpABEFGH (1.6 to 2.2), anaerobic nitrogen metabolism gene family napABC (3.5 to 5.3), catalase gene katY (−2.42), and 44 genes encoding hypothetical proteins. Changes were also identified in metabolic pathway regulation genes, including a key regulator, arcA (1.52), as well as major phosphotransferase system (PTS) regulators ptsG (−2.14) and ptsIH (1.45 to 1.81).

Significantly differently expressed genes indicating an attenuated phenotype. Download Table S3, PDF file, 0.8 MB.

An earlier study in which luxS was deleted in a Y. pestis Δpgm strain showed decreased hydrogen peroxide resistance, with downregulation in the expression of the katY gene, compared to the hydrogen peroxide resistance of the parental strain (7). Since we also observed downregulation in the expression of the katY gene in association with attenuation, a hydrogen peroxide resistance assay was performed, and we found minimal decreases in the ΔrbsA or the ΔlsrA strains but a significant decrease in hydrogen peroxide resistance in the attenuated ΔrbsA ΔlsrA mutant strain (see Fig. S2A in the supplemental material), consistent with the decreased expression of katY (see Table S3).

Functional assays of significant expression changes as determined by RNA-seq. (A) Resistance to killing by hydrogen peroxide. Resistance of indicated bacterial strains to killing by medium containing 0.3% H2O2 was evaluated through measurement of luminescence reporter in real-time. Measurements were taken 15 min after the addition of hydrogen peroxide. Statistical analysis was performed via one-way ANOVA with Tukey post hoc correction. (B) Growth of Y. pestis strains in carbon source-restricted minimal medium, measured by absorbance. Y. pestis strains were inoculated into a modified M9 medium, and samples were measured for absorbance at a wavelength of 600 nm every hour. Asterisks denote time points at which results for WT CO92 and the mutant were statistically significantly different at a P value of <0.001. Statistical analysis was performed using multiple t tests, with the Holm-Sidak method to correct for multiple comparisons. (C) Growth of Y. pestis strains in iron source-restricted minimal medium. Y. pestis strains were inoculated into a modified M9 medium, and samples were measured for absorbance at a wavelength of 600 nm every hour. Asterisks denote time points at which the results for WT CO92 and mutants were statistically significantly different at a P value of <0.001. Statistical analysis was performed using multiple t tests, with the Holm-Sidak method to correct for multiple comparisons. Data (arithmetic means ± standard deviations) are from three independent experiments. Download Figure S2, TIF file, 0.2 MB.

The changes in metabolic regulator gene expression were also especially intriguing given the significant attenuation observed for the growth of Y. pestis in the macrophages in nutrient-limited environments, as well as in the in vivo mouse model. To determine whether changes in metabolic gene expression could be altering the growth pattern of the ΔrbsA ΔlsrA mutant in a restricted-nutrient environment, the mutant and the WT CO92 strain were grown in a modified defined medium based on M9 salts. The ΔrbsA ΔlsrA mutant exhibited delayed growth kinetics when glucose was used as the primary carbon source (see Fig. S2B in the supplemental material). There was an extended lag phase of growth for the mutant, although it reached a final optical density (OD) equivalent to that of WT CO92.

Phenotype-suppressing gene expression was identified by sorting for unique changes between the ΔrbsA ΔlsrA mutant and WT CO92 that were also not evident between the ΔrbsA ΔlsrA ΔluxS mutant and WT CO92 (see Table S4 in the supplemental material). Through this analysis, 220 genes with a Padj of <0.1 were selected, which included a large segment of the genes encoding the type III secretion system (T3SS), including structural genes yscABCDGLOPRSTUVXY (1.3 to 2.6) and effector protein genes yopBDHJMQRT (1.4 to 2.9); all of these genes were upregulated, indicating globally increased expression of the T3SS genes.

Significantly differently expressed genes indicating suppression of attenuation. Download Table S4, PDF file, 0.8 MB.

The T3SS in Y. pestis has been extensively characterized as an essential virulence system with functions ranging from targeted cell lysis to immune evasion (28). We confirmed changes in the expression profiles of T3SS effectors, i.e., Yersinia outer protein E (YopE) and the structural low calcium response V antigen (LcrV), by Western blot analysis in luxS-associated and ΔrbsA ΔlsrA mutants, as well as WT CO92. Both of these proteins were secreted at higher levels in ΔluxS and ΔrbsA ΔlsrA ΔluxS mutants than in either WT CO92 or the ΔrbsA ΔlsrA mutant under inducing growth conditions in vitro, i.e., low calcium and 37°C (see Fig. S3A and B in the supplemental material) (28). Other virulence factor-encoding genes characterized, such as ail, the attachment and invasion locus, and pla, encoding plasminogen activator protease, were not significantly differentially expressed in any of the strains examined compared to their expression in WT CO92. We examined the levels of Pla by Western blot analysis (see Fig. S4B) and evaluated Pla protease activity (see Fig. S4A), and we found no significant difference among all mutants examined compared to the Pla level and activity of WT CO92.

Type III secretion system function. (A) Western blots of anti-LcrV and anti-YopE antibodies from concentrated culture supernatants normalized against total protein visualized on the blot. (B and C) Densities of anti-YopE (B) and anti-LcrV (C) antibodies were measured and plotted. Western blot image is representative of three independent experiments. Arithmetic means ± standard deviations are shown. Download Figure S3, TIF file, 0.8 MB.

Pla protease activities and levels. (A) The activity of Pla protease was measured for several strains using a fluorescent reporter assay. (B) The protein levels of Pla were determined by immunoblotting using polyclonal anti-Pla antibodies and with loading controlled by total protein visualization using stain-free technology (Bio-Rad). Data are from three independent cultures per strain. Arithmetic means ± standard deviations are shown. Download Figure S4, TIF file, 0.3 MB.

Finally, we identified changes in gene expression that were common to the ΔrbsA ΔlsrA and the ΔluxS mutant versus WT CO92 (see Table S5 in the supplemental material). This analysis identified 348 differentially expressed genes with a Padj of <0.1 and included several iron transport gene families, as well as AI-1 quorum-sensing components. There were significant changes in the expression of iron transport-related genes. Inorganic chelated iron transport genes yfeABCDE (2.15 to 3.75) were upregulated, while the organic iron transport gene tonB (−2.8) and siderophore yersiniabactin synthesis genes irp1 to -8 (−1.83 to −3.29) were both uniformly repressed. Iron transport is essential for the full virulence of Y. pestis, as has been demonstrated through deletions of the pgm locus, which includes the yersiniabactin synthetic family, the irp genes (29).

Significantly differently expressed genes in response to autoinducer-2 dysregulation. Download Table S5, PDF file, 0.9 MB.

To confirm the physiologic effects of these expression changes, we evaluated the growth of mutants in restricted-nutrient medium with a sole iron source of nonchelated inorganic iron, FeSO4. Restriction of the iron source resulted in delayed growth of the ΔrbsA ΔlsrA ΔluxS and ΔluxS mutants compared to the growth of WT CO92, as well as a lower final bacterial density, indicating a functional consequence of expression changes (see Fig. S2C in the supplemental material). In addition to the alterations in iron uptake mechanisms that were observed across AI-2-perturbed strains, we also observed uniform upregulation of AI-1 system component-encoding genes ypeIR (4.14 to 4.35) and yspI (2.02), including synthetic genes for both of the acyl-homoserine lactones used by the AI-1 system, as well as the downstream receptor (see Table S5).

AI-2 concentration dependence of Y. pestis CO92 virulence.

The transportation defect we identified in the ΔrbsA ΔlsrA mutant appeared to decrease early uptake of AI-2 into the bacteria, thus decreasing the amount of AI-2 available intracellularly. While the lack of AI-2 resulted in virulence comparable to that of WT CO92 in vivo, as was seen with luxS mutants, we were interested in the potential role of intermediate levels of AI-2, as might be occurring with the ΔrbsA ΔlsrA mutant strain. As has been reported previously by Rickard et al., responses to AI-2 can be dose dependent and bacteria may be sensitive to much lower levels of AI-2 than previously appreciated (30). Thus, we chemically rescued the ΔluxS strain with various concentrations of exogenous AI-2, ranging from 0 to the physiological level of AI-2, which we defined as the maximum concentration of AI-2 (2.5 µM) observed in the WT CO92 strain (Fig. 6). When we used quantitative real-time PCR (qRT-PCR) to analyze the expression of several key genes identified by RNA-seq (Fig. 5D), we found that the luxS deletion mutant with subphysiologic levels of AI-2 (0.25 µM) had decreased expression of katY, tonB, dnaK, and yopH (Fig. 5E), resembling the expression pattern of the ΔrbsA ΔlsrA mutant (Fig. 5D). Lack of AI-2 supplementation resulted in an expression profile with trends similar to those observed for the luxS mutant using RNA-seq (Fig. 5D and E), confirming the validity of those measures. Physiologic levels of AI-2 (2.5 µM) resulted in a gene expression profile in the luxS mutant that was similar to that of WT CO92 based on qRT-PCR (Fig. 5E).

FIG 6 

In vivo complementation of the ΔrbsA ΔlsrA ΔluxS mutant with exogenous synthetic AI-2. (A) Survival of female Swiss-Webster mice (n = 5) in a pneumonic plague model after challenge with 10-LD50 equivalent of WT CO92, where 1 LD50 is 500 CFU. Mice were dosed with the indicated concentrations of synthetic AI-2 in PBS via the intranasal (i.n.) route at the time of infection and at 24 h and 48 h postinfection. Data are representative of three independent experiments. Statistical analysis was performed using Kaplan-Meier survival curve analysis.

Finally, we attempted to rescue phenotypes in vivo with exogenous AI-2. In the mouse pneumonic plague model, we partially rescued the ΔrbsA ΔlsrA ΔluxS mutant with exogenous AI-2 to an attenuated phenotype at doses calculated to achieve concentrations of 0.2 and 2 µM in the lungs of mice (Fig. 6A and B). Mice that were infected with the WT CO92 strain and received equivalent doses of AI-2 showed no difference in virulence (Fig. 6A), suggesting that the change in virulence is specific to the ΔrbsA ΔlsrA ΔluxS strain and due to the availability of exogenous AI-2. There was significantly greater attenuation in a group of mice infected with the ΔrbsA ΔlsrA ΔluxS mutant and receiving 0.2 µM of AI-2 than in animals receiving no exogenous AI-2 (P = 0.0154). Increasing the exogenous concentration of AI-2 to 25 µM did not provide any protection to mice when challenged with the ΔrbsA ΔlsrA ΔluxS mutant (Fig. 6A). Additionally, the ΔrbsA ΔlsrA ΔluxS ΔlsrK mutant strain, which was insensitive to AI-2, could not be rescued by the addition of AI-2 and was fully virulent regardless of the dose of exogenous AI-2 provided (Fig. 6B).

DISCUSSION

The autoinducer-2 system has been proposed to be an interspecies metabolic-status signaling mechanism in bacteria, allowing adaptive regulation in response to environmental conditions. AI-2 controls a diverse array of traits in both nonpathogenic and pathogenic bacteria. In a ΔrbsA ΔlsrA mutant strain, we showed that aberrations in the AI-2 signaling mechanism resulted in a drastic reduction in virulence, a more-than-50-fold change as determined by LD50, compared to the virulence of the WT CO92 strain. This contrasts greatly with previous reports of AI-2 regulation both in Y. pestis and in other pathogenic bacteria (7, 10). We also demonstrated a basis for the reported differences observed in the context of AI-2 in virulence in vivo through deletion of the luxS gene and observation of its suppression of ΔrbsA ΔlsrA attenuation.

As has often been discussed in the literature, the role of LuxS in the cycling of homocysteine could have significant effects on the metabolic activity of bacteria. Furthermore, the loss of AI-2 production due to luxS deletion or, alternately, an endogenous AI-2 signal propagation that is distinct from exogenous signal propagation could play an important role in bacterial virulence (31). However, we showed that deletion of the lsrK gene from WT CO92 also resulted in a similar suppressive phenotype in a background of mutant (ΔrbsA ΔlsrA) attenuation, like the luxS deletion. LsrK phosphorylates AI-2, which binds LsrR, a repressor, inactivating it to trigger expression of the lsr operon (21, 27). These results suggest that the suppression of bacterial attenuation is due to the lack of AI-2 activity rather than the pleiotropic consequences of luxS deletion. Interestingly, the suppressive phenotype seems to result not from reversion to WT CO92 gene expression patterns but, rather, from upregulation of the T3SS of Y. pestis. This is not an isolated phenomenon either, as the upregulation of T3SS observed in this study due to the lack of AI-2 activity parallels similar secretion phenotypes in both Aeromonas and Salmonella ΔluxS strains (4, 32). The dose dependence of gene expression profiles, as well as rescue of the ΔrbsA ΔlsrA ΔluxS mutant from lethal effects in an in vivo model with subphysiological concentrations of AI-2, reveal a new mechanism of AI-2 signaling. Our data support a dose-dependent signaling model in which the response to AI-2 is stratified into three categories: zero activity, low activity, and high activity. Much of the previous literature has focused on the zero- and high-activity categories of AI-2, and, based upon this model, may be missing the significant intermediate category in which we observed the greatest disruption in virulence. We suggest that, in light of our results, AI-2 signaling may require reevaluation in other well-studied bacterial pathogens to characterize phenotypes derived from subphysiologic concentrations of AI-2 rather than complete absence of AI-2.

The results obtained through RNA-seq contain an incredible density of data that may point to mechanisms of attenuation and the role of AI-2 in bacterial systems. The metabolic regulation observed with AI-2 perturbation, especially the PTS system, indicates a strong role for AI-2 in the adaptive response to different environmental niches. The PTS system regulates preferred sugar uptake and depends on the flux of sugars to balance the phosphorylation state of two major regulatory kinases, PtsIH, that influence both the transcription and function of non-PTS transporters. The diminished expression of ptsG could influence the phosphorylation states of these kinases, in addition to the expression changes identified for ptsIH, allowing further dysregulation. Previous studies have indicated that PtsI is essential for the uptake of AI-2 via a regulatory function (33), and thus, the reciprocal changes in the expression of PTS regulators suggest a complex and strictly controlled dynamic that utilizes AI-2 as a powerful environmental signal. Taken together, the attenuated virulence phenotype of a Y. pestis CO92 mutant (e.g., the ΔrbsA ΔlsrA mutant) that has a diminished ability to transport AI-2 and the modulation of gene expression suggest a decoupling of metabolic status from regulatory control, resulting in a maladaptive metabolic and stress response profile.

Finally, the high conservation of AI-2 transport mechanisms and signaling pathways in microbes presents a significant opportunity for small-molecule intervention (34). Current inhibitors of AI-2 have unknown activity in in vivo models of disease. The paucity of in vivo data, in conjunction with the lack of attenuation previously observed for luxS deletion strains of several pathogenic bacteria, have prevented further research into the utility of AI-2 inhibitors. The inhibitors characterized thus far and the development of potential drug targets in the RbsBAC and LsrABCD families of proteins represent an untapped resource in the fight against antibacterial resistance.

MATERIALS AND METHODS

Bacterial strains, plasmids, and cell culture.

The bacterial strains used in this study are described in Table 1. Y. pestis strains were cultured overnight at 28°C, unless specifically noted otherwise in the figure legends, with shaking at 180 rpm in heart infusion broth (HIB) (Difco; Voigt Global Distribution, Inc., Lawrence, KS) or grown for 48 h on 5% sheep blood agar (SBA) (Teknova, Hollister, CA) or HIB agar plates. As appropriate, the organisms were cultivated in the presence of antibiotics ampicillin, kanamycin, and polymyxin B at concentrations of 100, 50, and 35 µg/ml, respectively. All of the experiments with Y. pestis were performed in the Centers for Disease Control and Prevention (CDC)-approved select agent laboratory in the Galveston National Laboratory (GNL), University of Texas Medical Branch (UTMB).

TABLE 1 

Bacterial strains used in this study

Genotype	Description	Origin
WT CO92	Virulent Y. pestis biovar orientalis strain isolated in 1992 from a fatal human pneumonic plague case and naturally resistant to polymyxin B	CDC
ΔrbsA	rbsA deletion mutant of Y. pestis CO92	11
Δlpp ΔrbsA	lpp rbsA double deletion mutant of Y. pestis CO92	11
Δlpp ΔmsbB ΔrbsA	lpp msbB rbsA triple deletion mutant of Y. pestis CO92	11
ΔlsrA	lsrA deletion mutant of Y. pestis CO92	This study
ΔrbsA ΔlsrA	rbsA lsrA double deletion mutant of Y. pestis CO92	This study
ΔluxS	luxS deletion mutant of Y. pestis CO92	This study
ΔrbsA ΔluxS	rbsA luxS double deletion mutant of Y. pestis CO92	This study
ΔlsrA ΔluxS	lsrA luxS double deletion mutant of Y. pestis CO92	This study
ΔrbsA ΔlsrA ΔluxS	rbsA lsrA luxS triple deletion mutant of Y. pestis CO92	This study
ΔrbsA ΔlsrA(pBR-luxS)	rbsA lsrA double deletion mutant of Y. pestis CO92 transformed with pBR-luxS (Tcs)	This study
ΔrbsA ΔlsrA ΔluxS(pBR-luxS)	rbsA lsrA luxS triple deletion mutant of Y. pestis CO92 transformed with pBR-luxS (Tcs)	This study
ΔrbsA ΔlsrA ΔluxS::Tn7-luxS	rbsA lsrA luxS triple deletion mutant of Y. pestis CO92 integrated with a Tn7 cassette carrying the native promoter and gene for luxS	This study
ΔrbsA ΔlsrA::Tn7-rbsA	rbsA lsrA double deletion mutant of Y. pestis CO92 integrated with a Tn7 cassette carrying the native promoter and gene for rbsA	This study
ΔrbsA ΔlsrA::Tn7-lsrA	rbsA lsrA double deletion mutant of Y. pestis CO92 integrated with a Tn7 cassette carrying the native promoter and gene for lsrA	This study
WT CO92::Tn7-lux	Y. pestis CO92 integrated with a Tn7 cassette carrying the luciferase operon luxCDABE (lux) (Kms)	This study
ΔrbsA::Tn7-lux	rbsA deletion mutant of Y. pestis CO92 integrated with a Tn7 cassette carrying the luciferase operon luxCDABE (lux) (Kms)	This study
Δlpp ΔrbsA::Tn7-lux	lpp rbsA double deletion mutant of Y. pestis CO92 integrated with a Tn7 cassette carrying the luciferase operon luxCDABE (lux) (Kms)	This study
Δlpp ΔmsbB ΔrbsA::Tn7-lux	lpp msbB rbsA triple deletion mutant of Y. pestis CO92 integrated with a Tn7 cassette carrying the luciferase operon luxCDABE (lux) (Kms)	This study
ΔlsrA::Tn7-lux	lsrA deletion mutant of Y. pestis CO92 integrated with a Tn7 cassette carrying the luciferase operon luxCDABE (lux) (Kms)	This study
ΔrbsA ΔlsrA::Tn7-lux	rbsA lsrA double deletion mutant of Y. pestis CO92 integrated with a Tn7 cassette carrying the luciferase operon luxCDABE (lux) (Kms)	This study
ΔluxS::Tn7-lux	luxS deletion mutant of Y. pestis CO92 integrated with a Tn7 cassette carrying the luciferase operon luxCDABE (lux) (Kms)	This study
ΔrbsA ΔluxS::Tn7-lux	rbsA luxS double deletion mutant of Y. pestis CO92 integrated with a Tn7 cassette carrying the luciferase operon luxCDABE (lux) (Kms)	This study
ΔlsrA ΔluxS::Tn7-lux	lsrA luxS double deletion mutant of Y. pestis CO92 integrated with a Tn7 cassette carrying the luciferase operon luxCDABE (lux) (Kms)	This study
ΔrbsA ΔlsrA ΔluxS::Tn7-lux	rbsA lsrA luxS triple deletion mutant of Y. pestis CO92 integrated with a Tn7 cassette carrying the luciferase operon luxCDABE (lux) (Kms)	This study
ΔlsrK	lsrK deletion mutant of Y. pestis CO92	This study
ΔrbsA ΔlsrA ΔlsrK	rbsA lsrA lsrK triple deletion mutant of Y. pestis CO92	This study
ΔrbsA ΔlsrA ΔluxS ΔlsrK	rbsA lsrA luxS lsrK quadruple deletion mutant of Y. pestis CO92	This study

RAW 264.7 murine macrophage cell lines (ATCC, Manassas, VA) were maintained in Dulbecco modified Eagle medium (DMEM) with 10% fetal bovine serum supplemented with 1% l-glutamine (Cellgro, Manassas, VA) and 1% penicillin-streptomycin (Invitrogen, Carlsbad, CA) at 37°C with 5% CO2.

Construction of flippase expression plasmid.

An easily curable plasmid for the expression of flippase recombinase, pEF01, was constructed on a pCP20 backbone that incorporated a levansucrase (sacB) gene derived from pDMS197 and lacked the chloramphenicol resistance cassette. The construction of the plasmid was accomplished using In-Fusion cloning (Clontech Laboratories, Inc., Mountain View, CA) with double-stranded DNA (dsDNA) fragments generated by PCR with primers P1 to P6 (Table 2).

TABLE 2 

Description of primers used in study

Primer	Description	Sequence (5′→3′)
P1	FLP, ampicillin, forward	CGCCTGTAGTGCCATTTACC
P2	FLP, ampicillin, reverse	CATTACGCTGACTTGACGGG
P3	FLP, forward	CCCGTCAAGTCAGCGTAATG
P4	FLP, reverse	GGTAGCGTTGCCAATGATGT
P5	SacB, FLP, forward	AATGGCACTACAGGCGTGGGAATTCTGATCCTTTTTAACCCATCAC
P6	SacB, FLP, reverse	TCATTGGCAACGCTACCGCCATTTGCCTGCTTTTATATAGT
P7	λ Red, LsrA, forward	ATTTGTTCAGTCCCGTCAGTCAACATTGAGGGAGCGGAGGCAACATGCAAGTGTAGGCTGGAGCTGCTTC
P8	λ Red, LsrA, reverse	CCGGTTATTTTGGATGAATTTCAACATGTTGCCTCCGACGCACCATGTTCCGGGGATCCGTCGACC
P9	λ Red, LuxS, forward	TTAGAAAAATATGACTTTTTTATGAGGAGGTAACTAAATGCCATTATTGGGTGTAGGCTGGAGCTGCTTC
P10	λ Red, LuxS, reverse	CGCCTTTTATCATTCTCCTGCCTACTGATACTGAGCACTAAATATGCAATATTCCGGGGATCCGTCGACC
P11	LsrA, Tn7, forward	CCAACACTCGAGAGGGCAAATAGGGTGAGAATG
P12	LsrA, Tn7, reverse	TCCTTCGAATTCAGCCACTGCGTAATGAATGTTT
P13	LsrA, reverse, sequencing	ATCTATCACCCCAGACTGCC
P14	LsrA, forward, sequencing	CCATCACGCCGTTCATTGAA
P15	LuxS, pBR322/Tn7, forward	TCCTTCGAATTCGCTTTGAAGAGTATTTAGCGCT
P16	LuxS, Tn7, reverse	CCAACAGGTACCAGCTTTACTGAACCCCCAGCC
P17	LuxS, pBR322, reverse	CCAACAGTCGACAAAGCTTTACTGAACCCCCAGCC
P18	LuxS, forward, sequencing	CAGTTATCTGCAGAGCGCGA
P19	LuxS, reverse, sequencing	GACGCTTTAATCAGCGCCTT
P20	λ Red, LsrK, forward	AGGGTATTCAAGAGGAGCGCGCAATGAGTCAACTCGATACGACTACCCCAGTGTAGGCTGGAGCTGCTTC
P21	λ Red, LsrK, reverse	AATGAAGATAATCCATTTTAGAGGCAAGGAGCCTTCCAAAGAGACGTCGTTTCCGGGGATCCGTCGACCT
P22	LsrK, reverse, sequencing	CGCCTTTAATCCTGCATGCT
P23	LsrK, forward, sequencing	CAGATATTGCGGTCGTTGGG

Construction of in-frame deletion mutants.

To construct in-frame deletion mutants of Y. pestis CO92, the λ phage recombination system was used (35). Initially, the WT CO92 strain was transformed with plasmid pKD46 and grown in the presence of 1 mM l-arabinose to induce the expression of the λ phage recombination system. The above-mentioned Y. pestis culture was processed for the preparation of electroporation-competent cells (35, 36). The latter were then transformed with 0.5 to 1.0 µg of the linear dsDNA constructs carrying the kanamycin resistance (Kmr) gene cassette, which was immediately flanked by the bacterial flippase recognition target (FRT) sequence, followed on either side by 50 bp of DNA sequences homologous to the 5′ and 3′ ends of the gene to be deleted from WT CO92. Plasmid pKD46 was cured from the mutants that had successful Kmr gene cassette integration at the correct location by growing the bacteria at 37°C. The latter mutants were transformed with plasmid pEF01 to excise the Kmr gene cassette. Eventually, plasmid pEF01 was also cured from the kanamycin-sensitive (Kms) clones by growing them at 37°C, followed by selection in a medium containing 5% sucrose (37). To confirm the in-frame deletion, mutants showing sensitivity to kanamycin and ampicillin were tested by PCR using appropriate primer pairs (Table 2) and sequencing of the PCR products.

Growth curve and AI-2 determination of mutants.

To determine the AI-2 secretion profile, overnight cultures of bacteria were inoculated in HIB medium at a dilution of 1:1,000, and then aliquots of culture were taken at hourly intervals. The culture medium was centrifuged briefly and then filtered through 0.1-μm microcentrifuge filters (Corning Inc, Corning, NY) before being stored at −80°C prior to analysis. Analysis was performed as previously described (38); in brief, V. harveyi BB170 (ATCC), which is unable to synthesize AI-2, was inoculated into AB medium, incubated overnight at 30°C, and then diluted 1:5,000 in fresh AB medium (38). Freshly diluted V. harveyi BB170 was then mixed 9:1 with filtered culture supernatants of Y. pestis strains, and the mixture incubated for 5 h at 30°C. Samples were then analyzed for bioluminescence, and AI-2 concentrations determined by a standard curve obtained with synthetic AI-2 (Omm Scientific, Dallas, TX).

Development of luminescent reporter strains.

Electrocompetent cells of Y. pestis strains were prepared and electroporated with pTNS2 and pUC18r6kT mini-Tn7T::lux-FRT-kan (39) and selected by kanamycin resistance and luminescence. Following isolation, strains were electroporated with pEF01 to remove the resistance cassette. Kanamycin-sensitive mutants were grown at 37°C and selected for on medium containing 5% sucrose for removal of pEF01. The insertion of the lux (luciferase) operon at the attTn7 region and appropriate removal of the kanamycin cassette were confirmed by PCR and Sanger sequencing. The luminescence intensity of each strain was determined by serial dilution and relative luminescence unit (RLU) measurement (Spectramax M5e; Molecular Devices, Sunnyvale, CA).

Intracellular survival of Y. pestis CO92 strains in RAW 264.7 murine macrophages.

Intracellular survival of Y. pestis strains was determined as previously described (40). In brief, luminescent Y. pestis strains were grown in HIB overnight to saturation at 28°C. RAW 264.7 macrophages were seeded in 96-well plates at a concentration of 2 × 104 cells/well for confluence. Plates were then infected with Y. pestis CO92 lux or the various mutant strains, also with lux, at a multiplicity of infection (MOI) of 250 in DMEM, centrifuged, and incubated at 37°C and 5% CO2 for 60 min. Infected macrophages were then washed with phosphate-buffered saline (PBS), treated with gentamicin, washed again with PBS, and maintained in DMEM as described above. At 0 and 4 h, luminescence was measured in a SpectraMax M5e microplate reader.

Y. pestis CO92 pneumonic plague mouse model.

All of the animal studies with Y. pestis were performed in an animal biosafety level 3 (ABSL-3) facility under an approved Institutional Animal Care and Use Committee (IACUC) protocol (UTMB). Six- to 8-week-old female Swiss Webster mice (17 to 20 g), purchased from Taconic Laboratories (Germantown, NY), were anesthetized by the intraperitoneal route with a mixture of ketamine and xylazine and subsequently challenged intranasally with the indicated (as shown in the figures) LD50 doses (1 LD50 = 500 CFU) as described for WT Y. pestis CO92 (41). Mice were assessed for morbidity and/or mortality, as well as clinical symptoms, for the duration of each experiment (up to 21 days postinfection).

For the AI-2 complementation study, mice were anesthetized by isoflurane and dosed intranasally at the time of infection and at 24 h and 48 h postinfection with 20 μl of PBS with added synthetic AI-2 calculated to result in a 0, 0.2, 2, or 25 μM concentration of AI-2 in the lung volume of a 6- to 8-week-old female Swiss Webster mouse (~500 μl).

AI-2 uptake by Y. pestis.

AI-2 uptake was determined as previously described (33). In brief, strains were grown to saturation overnight in HIB at 37°C. Bacteria were washed twice with PBS and diluted 1:100 in fresh HIB supplemented with 50 μM AI-2. Culture aliquots were sampled and assayed for AI-2 levels as described above.

Hydrogen peroxide resistance of Y. pestis strains.

Luminescent reporter Y. pestis strains were cultured as described for AI-2 analysis, but bacteria were harvested at the time of maximal AI-2 production, washed twice in PBS, and resuspended at an optical density at 600 nm (OD600) of 1 in HIB supplemented with 0.3% H2O2 (Thermo Fisher Scientific, Waltham, MA). Luminescence was measured in a SpectraMax M5e microplate reader.

Growth curve in modified minimal medium.

Overnight cultures of various Y. pestis strains were washed in PBS and then normalized by OD600. Flasks containing 20 ml of modified M9 medium (1× M9 salts [22 mM KH2PO4, 33.7 mM Na2HPO4, 8.55 mM NaCl, 9.35 mM NH4Cl], 1 mM MgSO4, 2.5 mM CaCl2, 0.001 mg/ml FeSO4, 0.0001% thiamine, 0.1% Casamino acids; all chemicals obtained from Sigma-Aldrich, St. Louis, MO) were supplemented with 0.4% glucose, inoculated with approximately 1 × 107 CFU of various bacterial strains, and incubated at 37°C with shaking at 180 rpm. Samples were taken every hour, and absorbance measured at OD600.

RNA-seq and expression analysis.

Cultures were grown as described for AI-2 growth curve analysis, and RNA was isolated at the time of peak AI-2 levels by using TRIzol (Thermo Fisher Scientific, Waltham, MA) and extracting with chloroform and ethanol. Total RNA was purified and DNase treated using the Quick-RNA kit (Zymo Research, Irvine, CA), followed by mRNA enrichment using MicrobeExpress (Ambion, Thermo Fisher Scientific, Waltham, MA).

Library construction and sequencing.

RNA (1 to 3 µg) was fragmented by incubation at 94°C for 8 min in 19.5 µl of fragmentation buffer per the manufacturer′s instructions (Illumina, San Diego, CA). Sequencing libraries were prepared using an Illumina TruSeq stranded-RNA kit, version 2, following the manufacturer’s protocol. The indexed samples were sequenced on a single lane of an Illumina HiSeq 1500 using the 2 × 50 paired-end protocol. The resulting BCL files were converted to fastq files using Illumina bcl2fastq2 software, version 2.17. Reads were checked for quality using FastQC and aligned to the Y. pestis CO92 genome using Burrows-Wheeler Aligner (BWA) via Illumina BaseSpace. Transcript counts were generated using the Bioconductor GenomicAlignments package (42), and differential expression was determined using DESeq2 (43). Euclidean distance mapping was performed using the distance function in R from the regularized-logarithm-transformed counts. Poisson distance mapping was performed using the method described by Witten (44), and heat maps were generated in R.

qRT-PCR.

Quantitative real-time PCR (qRT-PCR) was performed using the QuantiFast Sybr green PCR kit (Qiagen), according to the manufacturer’s protocol, and a 7300 real-time PCR system (Applied Biosystems, Grand Island, NY). Experimental gene mRNA levels were corrected to the level of 16S rRNA, and relative levels were calculated in relation to mRNA levels in the WT CO92 control. The results shown are the averages and standard deviations from three experiments. RNA samples were quantified using a NanoDrop spectrophotometer (NanoDrop Technologies, Wilmington, DE) and qualified by analysis on an RNA nano chip using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). Synthesis of cDNA was performed with 0.5 µg of total RNA in a 20-µl reaction mixture using the reagents in the TaqMan reverse transcription reagent kit from Applied Biosystems. The reaction conditions were as follow: 10 min at 25°C, 30 min at 48°C, and 5 min at 95°C. PCR amplifications were performed using 5 µl of cDNA in a total volume of 50 µl FailSafe buffer C (Epicenter Biotechnologies, Madison, WI).

Western blotting for T3SS effectors.

Western blotting for Y. pestis-secreted T3SS effectors was performed as previously described (40). In brief, overnight cultures of Y. pestis CO92 or its mutants, grown in HIB at 28°C, were diluted 1:20 in 5 ml HIB supplemented with 5 mM EGTA to trigger the low-calcium response. The cultures were incubated at 28°C for 2 h before being shifted to 37°C (to activate the T3SS) for an additional 3 h of growth. Supernatants were precipitated with 20% (vol/vol) trichloroacetic acid (TCA) on ice for 2 h. The TCA precipitates were then washed and dissolved in SDS-PAGE buffer and analyzed by immunoblotting using antibodies to YopE or LcrV (Santa Cruz Biotechnology, Santa Cruz, CA). The secondary antibodies were anti-rabbit IgG or anti-mouse IgG as appropriate (Southern Biotech, Birmingham, AL). Blots were developed using SuperSignal West Dura (Pierce Biotechnology, Thermo Fisher Scientific, Waltham, MA). Protein-loading normalization was accomplished through visualization of total protein in blots using stain-free technology (Bio-Rad, Hercules, CA).

Pla protease activity.

Pla protease activity measurement was performed as previously described (40). In brief, bacteria were grown as described above for AI-2 analysis and collected at the time of maximal AI-2 level. Cultures were centrifuged, washed twice, and resuspended in PBS to obtain a final OD600 of 0.1 using a spectrophotometer (SmartSpec 300; Bio-Rad). For each sample, 50-μl suspensions were added to wells of a black microtiter plate (Costar Corning, Inc., Corning, NY) in triplicate. The hexapeptide substrate dabcyl-Arg-Arg-Ile-Asn-Arg-Glu-(edans)-NH2, synthesized on Sieber amide resin (45), was added to the wells at a final concentration of 2.5 μg/50 μl. The kinetics of substrate cleavage by Pla was measured every 10 min for 3 h by a fluorometric assay (excitation/emission wavelengths, 360/460 nm) at 37°C on a BioTek Synergy HT spectrophotometer (BioTek Instruments, Inc., Winooski, VT).