Droplet Tn‐Seq identifies the primary secretion mechanism for yersiniabactin in Yersinia pestis

Abstract

Nutritional immunity includes sequestration of transition metals from invading pathogens. Yersinia pestis overcomes nutritional immunity by secreting yersiniabactin to acquire iron and zinc during infection. While the mechanisms for yersiniabactin synthesis and import are well‐defined, those responsible for yersiniabactin secretion are unknown. Identification of this mechanism has been difficult because conventional mutagenesis approaches are unable to inhibit trans‐complementation by secreted factors between mutants. To overcome this obstacle, we utilized a technique called droplet Tn‐seq (dTn‐seq), which uses microfluidics to isolate individual transposon mutants in oil droplets, eliminating trans‐complementation between bacteria. Using this approach, we first demonstrated the applicability of dTn‐seq to identify genes with secreted functions. We then applied dTn‐seq to identify an AcrAB efflux system as required for growth in metal‐limited conditions. Finally, we showed this efflux system is the primary yersiniabactin secretion mechanism and required for virulence during bubonic and pneumonic plague. Together, these studies have revealed the yersiniabactin secretion mechanism that has eluded researchers for over 30 years and identified a potential therapeutic target for bacteria that use yersiniabactin for metal acquisition.

While yersiniabactin is an essential virulence factor for several pathogens, the mechanism responsible for its secretion has eluded researchers. Droplet TnSeq enabled the identification of an AcrAB efflux system necessary for yersiniabactin secretion.

Introduction

Yersinia pestis causes the human disease plague. Like all microbes, Y. pestis requires transition metals for proper cell function and metabolism. However, eukaryotic hosts use mechanisms referred to as nutritional immunity to restrict access to these metals (Kehl‐Fie & Skaar, 2010; Hood & Skaar, 2012; Skaar & Raffatellu, 2015; Lonergan & Skaar, 2019). Consequently, Y. pestis has developed efficient systems to acquire iron (Fe), zinc (Zn), and manganese (Mn) to overcome nutritional immunity (Bearden & Perry, 1999; Perry et al, 2007, 2012; Desrosiers et al, 2010; Fetherston et al, 2010, 2012; Bobrov et al, 2017). The most well‐studied metal acquisition system in Y. pestis is the secreted metallophore yersiniabactin (Ybt), which is essential for iron acquisition and contributes to zinc acquisition during infection (Bobrov et al, 2014; Price et al, 2021). The mechanisms involved in the synthesis and import of Ybt have been extensively studied (Bearden et al, 1997; Gehring et al, 1998; Fetherston et al, 1999; Bobrov et al, 2017). The genes involved in these processes are located in four operons within the pigmentation (pgm) locus on the chromosome of Y. pestis (Fig 1A) (Bearden et al, 1997). YbtA regulates the expression of the genes that comprise the Ybt enzyme complex responsible for the synthesis of Ybt (irp1, irp2, ybtE, ybtS, ybtT, and ybtU) and that encode the transporters responsible for the import of Ybt bound to Fe (psn, ybtP, and ybtQ) or Ybt bound to Zn (ybtX) (Fetherston et al, 1996, 1999; Bobrov et al, 2014). These genes are conserved in other species that synthesize Ybt, including pathogenic Escherichia coli, Klebsiella pneumoniae, and Proteus mirabilis (Schubert et al, 1998; Karch et al, 1999; Bach et al, 2000; Oelschlaeger et al, 2003; Himpsl et al, 2010). Despite our extensive knowledge of this system, the molecular mechanism(s) responsible for the export of Ybt has yet to be identified. Because Ybt is an essential virulence factor in all bacteria which synthesize it (Fetherston et al, 2010), identifying the Ybt export mechanism could reveal a target for the development of anti‐virulence‐based therapeutics.

Figure 1. Ybt loci and droplet Tn‐seq experimental design to identify Ybt secretion system.

1. The Ybt genomic island and protein functions.
2. The Y. pestis transposon mutant library was grown in batch or droplet culture in cPMH2 for 24 h. Purple dots represent Ybt secretion mutants.
3. Fitness scores for mutants in Ybt genes or the AcrAB efflux system (y0703 and y0704).
4. The AcrAB‐TolC efflux system operon consists of three genes, y0702 (acrA), y0703 (acrB), and y0704 (tolC).

One reason the Ybt export mechanism has eluded researchers for the last 30 years is that the mechanism is not encoded with the rest of the Ybt genes. Therefore, high‐throughput genome screening techniques like transposon insertion sequencing (Tn‐seq) are needed to identify these essentially unknown genes. Traditional Tn‐seq uses a transposon to generate libraries of random mutants that include the inactivation of every non‐essential gene in the genome (van Opijnen et al, 2009). These libraries are then grown under selective conditions and the fitness of individual transposon mutants can be calculated by quantifying their abundance within the population by deep sequencing. Genes required for growth under the selective condition with transposon insertions will be less fit and make up a smaller portion of the population after selection. While Tn‐seq is a powerful discovery tool, conventional Tn‐seq has a limitation that it is unable to identify genes for secreted factors. This is because all the mutants in a transposon library are grown together in one large pool. Within this pool, mutants in secreted factors required for growth, for example, Ybt needed by Y. pestis to grow under metal deplete conditions, can “cheat” and use the secreted factors produced by their neighbors (also referred to as trans‐complementation). Cheating allows these mutants to still grow under selective conditions, hiding their true phenotype, and eliminates the possibility of identifying genes of secreted factors via conventional Tn‐seq. To overcome this limitation, we applied a recently described modified Tn‐seq application called droplet Tn‐seq (dTn‐seq) (Thibault et al, 2019) to identify mutants deficient in Ybt secretion. dTn‐seq uses microfluidics to encapsulate individual transposon mutants inside a medium droplet surrounded by an oil layer. The oil layer inhibits diffusion of Ybt, and other secreted factors, between droplets, allowing for the screening of a large library of mutants for growth defects under metal‐limited conditions unaffected by neighboring bacterial cells. Using dTn‐seq, we identified a role for an AcrAB‐TolC efflux system for growth of Y. pestis in metal‐limited conditions. Furthermore, by using a combination of in vitro and in vivo methods, we finally identified this AcrAB‐TolC system as the primary exporter for Ybt and showed it is required for Y. pestis virulence.

Results and Discussion

dTn‐seq can overcome trans‐complementation and identify factors involved in Ybt synthesis

We previously demonstrated that Y. pestis can grow within the microenvironment of the droplets needed for dTn‐seq (Thibault et al, 2019). For the Tn‐seq library, we generated ~50,000 mutants using the HIMAR transposon in a Y. pestis pCD1⁽⁻⁾ znuBC background. This background was chosen for two reasons. First, because the pCD1 plasmid is required for virulence but not for Ybt export, we can perform dTn‐seq in this background at BSL‐2. Second, while a znuBC mutant can grow in metal‐limited conditions, a znuBC irp2 mutant (irp2 is required for Ybt synthesis) is extremely attenuated for growth in the same medium (Bobrov et al, 2014). Therefore, transposon mutants in genes required for the synthesis, import, or export of Ybt in a znuBC background should have significant fitness defects that can be easily differentiated from the znuBC mutant in droplets containing Zn‐limited medium. The library was grown in BHI, passaged in metal‐chelated PMH2 medium (cPMH2) to deplete intracellular metal, and then grown as a batch culture in cPMH2 or encapsulated into droplets of cPMH2 (Fig 1B). Genomic DNA was isolated from each culture condition, sequenced, and individual transposon mutant fitness was calculated (Thibault et al, 2019). To validate that droplet encapsulation can overcome trans‐complementation, we first determined fitness phenotypes for known components of Ybt synthesis and import. As expected, mutants with transposon insertions in ybtA, the master regulator of the Ybt system (Fetherston et al, 1996), or in ybtX, required for Ybt‐Zn import (Bobrov et al, 2017), had significant fitness defects in batch culture (Fig 1C). In contrast, transposon insertions into the Ybt synthesis genes irp1, irp2, ybtE, or ybtT had no fitness defect in the batch culture, because despite not being able to synthesize Ybt (or at significantly lower levels in the case of the ybtT mutant; Miller et al, 2010), these mutants can still acquire Zn using Ybt synthesized from neighboring bacteria. However, in the droplets, both the import and synthesis mutants show significant fitness defects (Fig 1C). These data validate that encapsulation prevents trans‐complementation of Ybt between bacteria and support the use of dTn‐seq to screen for genes involved in Ybt secretion.

dTn‐seq revealed that y0703 and y0704 are required for Y. pestis growth in metal‐limited medium

Seventy‐two transposon mutants had statistically significant fitness phenotypes within the droplets (Dataset EV1 and Appendix Table S1). From this list, y0703 and y0704 showed significant fitness defects in the droplet but not the batch culture and have homology to components of efflux pumps. Furthermore, y0703 and y0704 are in an operon with y0702, and have homology to RND efflux systems (Fig 1D). Because similar efflux systems have been implicated in the export of siderophores in Vibrio cholerae and E. coli (Horiyama & Nishino, 2014; Kunkle et al, 2017), we prioritized our subsequent efforts to determine if these genes contribute to Ybt export in Y. pestis.

To validate our dTn‐seq data, we generated an in‐frame deletion of the entire y0702, y0703, and y0704 operon in Y. pestis pCD1⁽⁻⁾ znuBC. The mutant was grown in cPMH2 and compared to a znuBC mutant. The znuBC y0702‐y0704 mutant was significantly attenuated for growth compared to the znuBC mutant (Fig 2A; P ≤ 0.0001) but was not as attenuated as a znuBC irp2 mutant, which cannot synthesize Ybt. Growth was restored to znuBC levels by genetically complementation with y0702‐y0704 (Fig 2A) or by Zn supplementation (Fig 2B). Because Ybt is also involved in Fe acquisition, we generated a y0702‐y0704 mutant in Y. pestis pCD1⁽⁻⁾ to independently evaluate growth during Fe limitation. Under these conditions, y0702‐y0704 was significantly attenuated for growth in Fe‐limited conditions (Fig 2C; P = 0.0034) and had the same phenotype as the irp2 mutant in both limited and excess Fe conditions. Together, these data support that Y0702‐Y0704 contributes to growth during both Zn and Fe limitation.

Figure 2. y0702‐y0704 contributes to growth during metal limitation.

1. Growth of indicated Y. pestis strains in zinc‐limited medium (cPMH2) (n = 3).
2. Growth of indicated Y. pestis strains in zinc‐replete medium (cPMH2 + 10 μM ZnCl₂) (n = 3).
3. Growth of indicated Y. pestis strains in iron‐limited (cPMH2 + 0 μM FeCl₂) or iron‐replete (cPMH2 + 10 μM FeCl₂) media (n = 3).
4. Growth of znuBC y0702‐y0703, znuBC y0704, or y0702‐y0703 complemented mutant (pBCSK + y0702‐70703) in zinc‐limited medium (cPMH2) (n = 3).

Data information: One‐way ANOVA with Dunnett test, *P ≤ 0.05, **P ≤ 0.001, ****P ≤ 0.0001. For (A), (B), and (D), growth of mutants was compared to znuBC. Data represent the mean ± SD of three biological replicates. For some points, error bars are too small to visualize on the graph.

Source data are available online for this figure.

The Y0702‐Y0704 RND efflux system consists of an AcrAB inner membrane transporter (Y0702‐Y0703) and a TolC outer membrane transporter (Y0704) (Fig 2A). To confirm that both components are required for growth under Zn limitation, single mutants were generated in each system. Both mutants grew at the same rate as the znuBC y0702‐y0704 mutant, supporting that both components are required for growth under Zn limitation (Fig 2D). However, there was a chance that the deletion of y0702‐y0703 could have a polar effect on y0704 and that only y0704 was required. To eliminate this possibility, we trans‐complemented the znuBC y0702‐y0703 mutant with y0702‐y0703 on a plasmid, which restored growth (Fig 2D), demonstrating there was no polar effect on y0704 and that both the AcrAB and TolC components contribute to growth in metal‐limited conditions.

Y0702‐Y0704 are required for Ybt secretion

Due to the decreased growth of a znuBC y0702‐y0704 mutant in Zn‐limited medium, we next tested if y0702‐y0704 is required for Ybt secretion. Since Ybt is secreted into the growth medium, strains secreting Ybt can trans‐complement the growth of a znuBC irp2 mutant during co‐culture in metal‐limited medium (Bobrov et al, 2014). Therefore, to determine if a y0702‐y0704 mutant secretes Ybt, we co‐cultured a bioluminescent znuBC irp2 mutant with either Y. pestis pCD1⁽⁻⁾, the irp2 mutant, or the y0702‐y0704 mutant and measured znuBC irp2 growth as a function of bioluminescence (Fig 3A). As expected, Y. pestis pCD1⁽⁻⁾ secreted enough Ybt to support the growth of znuBC irp2 in Zn‐limited medium (Fig 3B). However, the irp2 mutant, which cannot synthesize Ybt, did not support growth. Likewise, co‐culture with the y0702‐y0704 mutant was unable to trans‐complement the growth of znuBC irp2 (Fig 3B), supporting that the y0702‐y0704 mutant is deficient in Ybt secretion. Complementation of the y0702‐y0704 system restored the ability of the strain to support growth of a znuBC irp2 mutant (Fig 3B).

Figure 3. y0702‐y0704 is important for Ybt secretion.

1. Bioluminescent co‐culture experimental design.
2. Growth of znuBC irp2 pGENlux in co‐culture with indicated Y. pestis strains in zinc‐limited medium (cPMH2) (n = 3).
3. Absorbance of PAR at 500 nm incubated with indicated Y. pestis supernatants (n = 3).
4. OD₆₀₀ of bacterial cultures at 8 h for LC–MS (n = 3).
5. LC–MS analysis of Ybt in culture supernatants (note: both peaks are Ybt).
6. LC–MS quantification of Ybt in culture supernatants (n = 3).
7. OD₆₀₀ of bacterial cultures at 8 h harvested for GFP quantification (n = 3).
8. Quantification of GFP protein (ratio of GFP in supernatant to cell lysates) (n = 3).

Data information: For (B), Data represents the mean ± SD relative luminescent units (RLU) from three biological replicates. For some points, error bars are too small to visualize on the graph. One‐way ANOVA with Dunnett test compared to znuBC irp2 + Y. pestis; ****P ≤ 0.0001. For (C–G), One‐way ANOVA with Tukey test, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001; ns = not significant. Mean ± SD of three biological replicates.

Source data are available online for this figure.

Since Ybt binds to Zn (Behnsen et al, 2021), we can indirectly measure Ybt in the bacterial supernatant by measuring Ybt competition of Zn‐binding with 4‐(2‐pyridylazol)resorcinol (PAR) (Hunt et al, 1985). PAR changes color when bound to Zn (absorbs light at 410 nm when unbound and 500 nm when bound to Zn), and therefore, the amount of free Zn within a sample can be measured by the change in the PAR absorbance. If a Zn‐binding competitor is present, less Zn is available to bind to PAR, resulting in a decreased absorbance at 500 nm. To test whether the supernatant from various Y. pestis strains could compete for Zn binding with PAR, bacteria were grown under metal‐limited conditions and equal volumes of culture supernatant were harvested and filtered to remove bacteria. When compared to supernatant from Y. pestis pCD1⁽⁻⁾, there was significantly higher PAR absorbance at 500 nm in supernatants from the irp2 and y0702‐y0704 mutants (Fig 3C; P < 0.01), further supporting that the y0702‐y0704 mutant is deficient in Ybt secretion.

Finally, to directly measure the amount of Ybt secreted by the y0702‐y0704 mutant, we used liquid chromatography mass spectrometry (LC–MS). Bacteria were grown in metal‐limited medium for 8 h, the OD₆₀₀ of the cultures were measured (Fig 3D), and Ybt was extracted with ethyl acetate from 10 OD₆₀₀ equivalents of cell‐free supernatant and analyzed by LC–MS (Miller & DeMoll, 2011) (Fig 3E). Significantly less Ybt was present in the supernatant of the y0702‐y0704 mutant compared to Y. pestis pCD1⁽⁻⁾ (Fig 3F; P ≤ 0.001). However, there appeared to be residual Ybt detected in the supernatant of y0702‐y0704 above that of irp2 (Fig 3F). To determine if y0702‐y0704 was undergoing increased lysis due to intracellular Ybt toxicity, which could result in release of Ybt, we transformed Y. pestis pCD1⁽⁻⁾ and y0702‐y0704 with a plasmid expressing GFP and measured the concentration of GFP in the culture supernatants by Western blot. We observed no difference in the amounts of GFP between the supernatants (Fig 3G and H), supporting that lysis of y0702‐y0704 did not occur at a higher rate. However, these data do not eliminate the possibility that the residual Ybt in the medium is not a result of release from normal bacterial lysis. Together, these data demonstrate that y0702‐y0704 is deficient in Ybt secretion and support that the Y0702‐Y0703 AcrA‐BTolC efflux system contributes to Ybt export.

Y0702‐Y0704 are required for Y. pestis virulence

Ybt is required for Fe acquisition and contributes to the ability of Y. pestis to compete for Zn with the host protein calprotectin, and is thus essential for the development of lethal bubonic and pneumonic plague (Fetherston et al, 2010; Price et al, 2021). If Y0702‐Y0704 exports Ybt, then we expect it to be required for competition for metals with calprotectin and virulence. To determine if Y0702‐Y0704 contributes to fitness in the presence of calprotectin, the concentration of calprotectin to inhibit 50% bacterial growth (IC₅₀) was determined for Y. pestis pCD1⁽⁻⁾, irp2, znuBC irp2, or znuBC y0702‐y0704. The IC₅₀ of calprotectin was significantly lower for znuBC y0702‐y0704 compared to Y. pestis pCD1⁽⁻⁾ (Fig 4A; P ≤ 0.01), demonstrating that the mutant is less fit in the presence of metal restriction by calprotectin. Genetic complementation with y0702‐y0704 restored the ability of the mutant to grow in the presence of calprotectin. To determine if y0702‐y0704 is required for Y. pestis virulence, mice were infected with Y. pestis strains in which we restored the pCD1 plasmid (pCD1Ap) (Gong et al, 2001) via either subcutaneous injection or intranasal instillation to mimic bubonic and pneumonic plague, respectively. As expected, all mice succumbed to infection with Y. pestis or the znuBC mutant but survived infection with znuBC irp2 (Fig 4B and C). Mice infected with znuBC y0702‐y0704 were more resistant to lethal infection, with a 50% survival rate in the bubonic plague model (Fig 4B; P ≤ 0.001) and 20% survival rate in the pneumonic model (Fig 4C; P ≤ 0.0001). Moreover, there was a significant delay in the mean time to death in the pneumonic plague model for the znuBC y0702‐y0704 infected mice (3.75 days) compared to those infected with wild‐type (2.5 days; P ≤ 0.0001) or the znuBC mutant (3 days; P = 0.0022). Genetic complementation with y0702‐y0704 restored virulence in both models (Fig 4B and C). Together, these data show that y0702‐y0704 contribute to virulence during bubonic and pneumonic plague.

Figure 4. Y0702‐Y0704 contribute to Y. pestis virulence.

1. IC₅₀ of calprotectin for indicated Y. pestis strains (n = 3).
2. C57BL6/J mice (n = 10) subcutaneously challenged with 10² CFU of indicated Y. pestis strains.
3. C57BL6/J mice (n = 10) intranasally challenged with 10⁴ CFU of indicated Y. pestis strains.
4. Bacterial bioluminescence in the lungs of intranasally infected mice (n = 10; mean ± SD).
5. Representative mice from (D).

Data information: For (A), mean ± SD of three biological replicates. One‐way ANOVA with Tukey's comparison: ns = not significant, **P < 0.01, ***P = 0.001. For (B) and (C), differences in survival were calculated by log rank test compared to Y. pestis. For (B–D), results are combined data from two independent experiments (n = 5 mice per experiment).

Source data are available online for this figure.

Previous studies have shown that Y. pestis unable to synthesize Ybt administered intranasally do not proliferate within the lungs but can cause lethal infection via dissemination to the blood where metal availability is greater (Bearden et al, 1997). Therefore, we next investigated whether znuBC y0702‐y0704 was defective in lung proliferation during pneumonic plague using a bioluminescent reporter (pGEN‐luxCDABE) and optical imaging to monitor znuBC y0702‐y0704 proliferation during infection (Sun et al, 2012). Similar to Y. pestis, the znuBC y0702‐y0704 mutant was able to replicate efficiently in the lungs of mice that developed lethal disease (Fig 4D and E), indicating that mice that succumbed to infection were developing pneumonic plague.

Other AcrAB systems do not contribute to Ybt secretion

These data support that Y0702‐Y0704 exports Ybt in Y. pestis, but the residual Ybt in culture supernatants and increased virulence of znuBC y0702‐0704 compared to znuBC irp2 also indicate the presence of a secondary Ybt export system. Genome analysis revealed three other AcrAB systems: y1049‐1050, y3760‐3759, and y3392‐3393 (Stirrett et al, 2008). To determine if these systems are capable to export Ybt, we generated in frame deletions in all three AcrAB transporters in the znuBC or znuBC y0702‐0704 backgrounds and determined the growth of the mutants in Zn‐limited medium. Deletion of all three AcrAB systems had no significant impact on the growth of either background (Fig 5), indicating that Y0702‐Y0703 is the only AcrAB efflux system capable of exporting Ybt in Y. pestis.

Figure 5. Other AcrAB transporters do not contribute to Ybt secretion.

Growth of indicated Y. pestis strains in zinc‐limited medium (cPMH2) (n = 3). One‐way ANOVA with Dunnett test compared to znuBC mutant, **P ≤ 0.01, ****P ≤ 0.0001. Data represents the mean ± SD of three biological replicates. For some points, error bars are too small to visualize on the graph. Source data are available online for this figure.

Implications and future directions

Thirty years after the first reports of Ybt as a siderophore produced by Y. enterocolitica (Haag et al, 1993), we have finally identified an AcrAB‐TolC efflux system as the primary Ybt export mechanism in Y. pestis. We were able to accomplish this through the application of dTn‐seq, which is uniquely suited to screen for mutations in secretion systems or factors that cannot be investigated with conventional Tn‐seq approaches that suffer from trans‐complementation hurdles. The successful application of dTn‐seq in the identification of Y0702‐Y0704 establishes the utility of this system to be used by others to identify export systems or secreted factors produced by a variety of bacteria. However, considerations are needed when establishing the experimental design for dTn‐seq. First, care needs to be taken to ensure a high frequency of the droplets produced contain a single bacterium. This frequency can be influenced by the aggregative nature of the bacteria, the concentration of the culture, and the flow rate of the microfluidics system. Even under the best optimized parameters, a certain population of droplets will contain more than one bacterium, which can diminish absolute fitness phenotypes in mutants that can cheat. This is likely the reason that we still recovered some Ybt biosynthesis mutants in the droplets in our screen (Fig 1C). Second, the need to maximize single bacterium encapsulation will also influence library size. While conventional Tn‐seq often employs transposon libraries ≥ 10⁶ mutants to ensure genome saturation, current microfluidics technology needed to provide an appropriate frequency of droplets with individual bacterium limits individual libraries to approximately 5–7 × 10⁴ mutants. While size of the individual pool cannot currently be increased, an alternative approach is to repeat the screen with multiple pools. Finally, the microenvironment of the droplet can affect bacteria differently than similar large volume culture conditions. Therefore, optimization of growth conditions in the droplets needs to be performed for each bacterial species and selective condition prior to performing screens. Despite these limitations, our success using this system to identify Y0702‐Y0704 as a Ybt export system demonstrate that dTn‐seq is a powerful new technique to identify secreted factors or systems that cannot be identified by conventional transposon mutagenesis strategies.

These studies demonstrated that the Y0702‐Y0704 AcrAB‐TolC efflux pump is important for growth in metal‐limited medium and Ybt secretion by Y. pestis. However, znuBC y0702‐y0704 was not as attenuated as znuBC irp2 in Zn‐limited medium and we were still able to recover Ybt from y0702‐y0704 cultures. While these Ybt concentrations were not sufficient for optimal growth during metal limitation or to trans‐complement the growth of znuBC irp2, it does indicate that some amount of extracellular Ybt is available for the bacteria. One possible source of Ybt in y0702‐0704 cultures could be Ybt released from dying bacteria. In E. coli and V. cholerae, inhibition of siderophore efflux leads to growth restriction, envelope stress, and induction of the Cpx stress response system but increased cell lysis was not reported (Kunkle et al, 2017; Guest et al, 2019). It is likely that inhibition of Ybt secretion would induce similar stresses in Y. pestis. However, we tried to indirectly compare bacterial lysis between y0702‐y0704 and Y. pestis by measuring the release of cytoplasmic expressed GFP and did not observe increased concentrations that would indicate elevated bacterial lysis of the mutant (Fig 4). While this does not rule out the possibility that normal rates of bacterial death could be responsible for the residual Ybt recovered in in vitro culture, znuBC y0702‐y0704 is still partially virulent in the animal model, where bacterial lysis seems less likely to be sufficient to sustain bacterial colonization of a small bacterial population. Therefore, it seems more likely that a secondary export system is responsible for the residual extracellular Ybt sufficient to support the intermediate virulence in the mouse model. Moreover, we also observed a more significant phenotype during bubonic plague than pneumonic plague. While this difference could be attributed to the inherently more stringent nature of the bubonic plague model, it is also possible that the secondary export system is differentially regulated within the two tissues. Such regulation would indicate tissue‐specific environmental signals sensed by the bacterium and a potential future research direction to better understand exporter redundancy once the secondary system is identified.

Our studies also eliminated the possibility that the secondary system is another AcrAB homolog within the Y. pestis genome. However, other RND efflux family proteins have been implicated in the efflux of other siderophores (Horiyama & Nishino, 2014; Henríquez et al, 2019), and multiple RND efflux homologs are present in the Y. pestis genome that are viable candidates for a secondary system (Stirrett et al, 2008). While none of these genes had fitness defects in the dTn‐seq screen, it is likely that these fitness defects were masked by Ybt exportation via the Y0702‐Y0704 system. Moving forward, we will continue to search for the secondary export system by specifically targeting these other RND homologs for inactivation or performing a second round of dTn‐seq in znuBC y0702‐y0704.

The Ybt operon is also in clinical isolates of E. coli, K. pneumoniae, P. mirabilis, and S. enterica serovars (Schubert et al, 1998; Karch et al, 1999; Bach et al, 2000; Oelschlaeger et al, 2003; Himpsl et al, 2010). Our discovery that an AcrAB‐TolC system is the primary exporter of Ybt in Y. pestis strongly indicates that AcrAB‐TolC efflux systems are the primary mechanisms for the export of Ybt in these other species. Because Ybt provides an important virulence advantage for these pathogens (Karch et al, 1999; Lawlor et al, 2007; Himpsl et al, 2010; Bachman et al, 2011; Aviv et al, 2014), a conserved export system makes for an attractive broad spectrum anti‐virulence drug target, especially since many of these bacteria are also rapidly developing resistance to conventual antibiotics. Importantly, AcrAB and other RND efflux systems are also effective mediators of antibiotic resistance (Ma et al, 1994; Nikaido et al, 1998; Riera et al, 2011; Kosmidis et al, 2012; Nowak et al, 2015; Lin et al, 2017) and have previously been targeted to develop drugs that inhibit antibiotic efflux (Aeschlimann et al, 1999; Kaatz et al, 2003; Schumacher et al, 2006; Torres et al, 2012; Adamson et al, 2015; Nzakizwanayo et al, 2017). Thus, there are already several drug classes that might be repurposed as anti‐virulence adjunct therapies with antibiotics to improve outcomes for infections by bacteria that use Ybt to acquire metals.

In conclusion, using dTn‐seq in Y. pestis, we have identified the AcrAB‐TolC efflux pump Y0702‐Y0704 as the primary secretion mechanism for Ybt. This successful utilization of dTn‐seq brings a new technique to the field to identify unknown secreted factors in bacteria. Importantly, these data support previous studies that implicate efflux pumps as important metallophore exporters for other bacteria (Horiyama & Nishino, 2014; Kunkle et al, 2017; Henríquez et al, 2019) and suggest a larger role for these pumps in bacterial pathogenesis and as potential drug targets.

Materials and Methods

Ethics statement

C57BL/6J mice were purchased from The Jackson Laboratory and bred within the barrier facility at the Clinical and Translational Research Building at the University of Louisville. Groups contained both male and female mice and we observed no differences in outcomes based on sex during these studies. Animals were housed in accordance with NIH guidelines and all procedures were approved by the University of Louisville IACUC (IACUC# 19524). Three days prior to challenge with Y. pestis, animals were transferred to the University of Louisville's Center for Predictive Medicine Regional Biocontainment Laboratory ABSL‐3 facilities for acclimation. After infection, mice were observed for up to 14 days for the development of moribund disease and humanely euthanized if they met predefined endpoints.

Bacterial strains

The bacterial strains used in this study are listed in Appendix Table S2. Y. pestis was routinely grown in Difco brain heart infusion (BHI) broth (BD Biosciences) or under metal deplete conditions in chelex‐treated PMH2 supplemented with 1.0 μM FeCl₂, 1.0 mM MgCl₂, and where appropriate, 0.6 μM ZnCl₂ (cPMH2) (Desrosiers et al, 2010). For metal replete experiments, cPMH2 was supplemented with either 10 μM ZnCl₂ or 10 μM FeCl₂ as needed. Strains transformed with the pGEN‐luxCDABE plasmid (Sun et al, 2012) were grown with carbenicillin (50 μg/ml) prior to co‐culture. For infection, bacteria were grown overnight at 26°C in BHI broth prior to subcutaneous challenge for the bubonic plague model. Prior to intranasal instillation, Y. pestis was diluted to 0.05 OD₆₀₀ in BHI broth with 2.5 mM CaCl₂ and grown at 37°C with aeration for 16–18 h (Sun et al, 2012). Bacterial concentrations were determined using a spectrophotometer and diluted to desired concentrations in 1× PBS for mouse infections. Concentrations of bacterial inoculums for mouse studies were confirmed by serial dilution and enumeration on agar plates.

Generation of transposon mutant library and site directed mutants

A transposon mutant library was generated in Y. pestis pCD1⁽⁻⁾ znuBC. To generate this library, we used the transposon delivery vector pSAM‐DKm encoding the mariner‐family transposon, Himar1 C9 transposase, as previously described (Gutierrez et al, 2015). Briefly, Y. pestis was transformed with pSAM‐DKm by conjugation with E. coli S17 λPIR and transposon mutants were selected on BHI agar plates with 50 μg/ml kanamycin (plasmid selection) and 2 μg/ml irgasan (counterselection against E. coli) and grown at 26°C for 2 days. Colonies were enumerated and combined from multiple plates to generate a library containing approximately 50,000 mutants (representing > 10× coverage of the genome). The library was cryopreserved after growing in BHI broth with 50 μg/ml kanamycin, 2 μg/ml irgasan, and 1 μM ZnCl₂ shaking at 26°C for 2 days. Random distribution of transposon insertions was confirmed by selecting 20 colonies before cryopreservation and comparing PCR products generated using nested PCR with transposon specific and random primers (Gutierrez et al, 2015). Site specific and complemented mutants were generated using pSR47s and homologous recombination as previously described (Merriam et al, 1997; Walker & Miller, 2004). For trans‐complementation, y0702‐y0703 was ligated into pBC‐SK⁺, as previously described (Al‐Khodor et al, 2008). Deletions were confirmed by PCR and sequencing.

Microfluidic device production for dTn‐seq

Microfluidic devices were produced following the protocol as previously described (Thibault et al, 2019). Briefly, a microfluidic device mask with a single aqueous inlet and a 40 × 40 μm channel at the flow‐focus junction was designed using AutoCad 2016 software. Glass photomasks were ordered from CAD/Art Services, Inc. (Bandon, OR). These masks were used to generate silicon molds by soft lithography that were used for the final microfluidic fabrication using polydimethylsiloxane (PDMS). The silicon mold and final microfluidic chip fabrication (including PDMS‐glass plasma bonding) were performed at the Integrated Sciences Cleanroom and Nanofabrication Facility at Boston College.

Droplet production and culturing in droplets

The agarose droplets were produced following the protocol as previously described (Thibault et al, 2019). Briefly, 1% Seaplaque agarose (Lonza – 50101) was heated until dissolved in cPMH2 (Desrosiers et al, 2010). The agarose solution was filtered (0.45 μm) and incubated at 37°C. Transposon libraries were added at a suspension of 7.2 × 10⁶ cell/ml and thoroughly vortexed. Syringe pump systems were incubated in a 37°C for 2 h to prewarm them prior to droplet production. For oil and aqueous phases, a syringe pump rate of 500 μl/h was used, and the droplets were collected at 30 min. Before culturing, agarose droplets were gelled at 4°C for 12 min with occasional shaking.

Sample preparation, sequencing, and fitness calculations

Genomic DNA (gDNA) was extracted from Y. pestis using the DNeasy Blood & Tissue Kit according to the manufacturer's guidelines (Qiagen). Illumina DNA sample preparation for Tn‐Seq was performed as previously described (Thibault et al, 2019). For cleanup, droplet DNA was cleaned using 10 μl of magnetic beads mixed with 20 μl PEG solution per sample followed by elution in 14.3 μl of dH₂O. Batch samples were cleaned and precipitated with the phenol/chloroform method and resuspended in 26 μl of dH₂O. The ligation of DNA adapter barcodes was performed using T4 DNA ligase (NEB M0202L). The mix was combined with 13.12 μl DNA (droplet DNA) or 25 μl DNA (batch DNA) to 1 μl of 1:5 diluted adapter, 1× T4 DNA Ligase Reaction Buffer, and 400 units T4 DNA ligase, followed by incubation at 16°C for 16 h, 65°C for 10 min, and held at 10°C. In the droplet DNA sample, 10 μl magnetic beads were mixed with 20 μl PEG solution followed by elution in 36 μl of dH₂O. Batch sample was used directly. Ligated DNA was PCR amplified using Q5 high‐fidelity DNA polymerase (NEB – M0491L) using 34 μl of DNA (droplet DNA) or 1 μl of ligation mix added to 1× Q5 reaction buffer, 10 mM dNTPs, 0.45 μM of each primer (P1‐M6‐GAT‐MmeI; P2‐ADPT‐Tnseq‐primer), and one unit of Q5 DNA polymerase (Thibault et al, 2019). The samples were incubated at 98°C for 30 s, then followed by 21 cycles of 98°C for 10 s, 62°C for 30 s, 72°C for 15 s, followed by 72°C for 2 min, and hold at 10°C. According to the manufacturer's protocol, PCR products were gel purified and sequenced on an Illumina NextSeq 500. Sequence analysis was performed as previously described (van Opijnen et al, 2009; McCoy et al, 2017; Anthony & van Opijnen, 2019; Thibault et al, 2019).

PAR assay

To test for competition for zinc binding with 4‐(2‐pyridylazo)resorcinol (PAR), bacteria were grown at 26°C in BHI broth for 15 h, diluted to 0.1 OD₆₀₀ in cPMH2, and grown with aeration at 37°C for 8 h. Each culture was passaged once more in cPMH2 for 15 h at 37°C. Cell‐free supernatant was isolated by filtration through a 0.02 μM filter to remove Y. pestis and incubated with 50 μM ZnCl₂ for 15 min. PAR was added at 200 μM and incubated for 5 min. Absorbance was read on a Biotek Synergy HT plate reader using a spectral read from 350 to 550 nm.

Co‐culture trans‐complementation to measure rescue of Y. pestis znuBC irp2 growth

Bacteria were grown at 26°C in BHI broth for 15 h, diluted to 0.1 OD₆₀₀ in cPMH2, and grown with aeration at 37°C for 8 h. Each culture was passaged once more in cPMH2 for 15 h at 37°C. Bacteria were resuspended into fresh cPMH2 at a final concentration of 1 × 10⁶ CFU/ml. One hundred microliter of Y. pestis znuBC irp2 carrying the pGEN‐luxCDABE bioluminescent reporter were aliquoted in to each well of a white 96‐well plate (Greiner Bio‐One). 100 μl of Y. pestis, irp2, or y0702‐y0704 were added to individual wells (n = 3) and the plates were incubated at 37°C. Growth of Y. pestis znuBC irp2 as a function of increased bioluminescence was measured every 2 h for 8 h using a Biotek Synergy HT plate reader (0.5 s read, sensitivity of 135) as previously described (Sun et al, 2012).

Measurement of Ybt concentrations by LC/MS

Bacteria were grown at 26°C in BHI broth for 15 h, diluted to 0.1 OD₆₀₀ in cPMH2, and grown with aeration at 37°C for 8 h. Each culture was passaged once more in cPMH2 for 15 h at 37°C. Bacteria were diluted to 0.1 OD₆₀₀ in cPMH2 and grown for 8 h at 37°C. Ten milliliter of cell‐free supernatants were collected by centrifugation and filtration and mixed at a 1:1 ratio with ethyl acetate for extraction as previously described (Miller & DeMoll, 2011). The extract was dissolved in methanol and diluted in water for HPLC analysis as previously described (Miller & DeMoll, 2011).

Cell lysis measurement

Yersinia pestis strains transformed with pGEN222::gfp (Galen et al, 1999) were grown at 26°C in BHI broth for 15 h. Bacteria were diluted to 0.1 OD₆₀₀ in cPMH2 and grown at 37°C for 8 h. The culture was passaged once more in cPMH2 and grown for 15 h at 37°C. Four OD₆₀₀ equivalents of cell pellets and cell‐free supernatants were collected. For supernatants, proteins were concentrated by trichloroacetic acid (TCA) precipitation (10% final concentration). Samples were incubated for 20 min on ice and proteins were pelleted at 16,200 g for 10 min. The proteins were washed with acetone, pelleted at 16,200 g for 2 min, and dried prior to resuspending directly in loading buffer. GFP was detected by slot blot using an anti‐GFP antibody (Sigma No. G1544; 1:5,000).

Determination of calprotectin inhibitory concentrations

Recombinant forms of calprotectin were produced in E. coli and purified as previously described (Kehl‐Fie et al, 2011; Damo et al, 2013). To determine the concentrations of calprotectin to inhibit 50% growth of Y. pestis, bacteria were incubated with increasing concentrations of calprotectin and bacterial growth was determined as previously described (Price et al, 2021).

Animal infections with Y. pestis

Mice were challenged with Y. pestis as previously described (Sun et al, 2012; Dinc et al, 2014; Bowen et al, 2019). For intranasal challenge, mice were anesthetized with ketamine/xylazene and administered 20 μl of bacteria suspended in 1× PBS to the left nare. For subcutaneous challenge, mice were anesthetized with isoflurane and administered 20 μl of bacteria suspended in 1× PBS via subcutaneous injection at the base of the tail. Infected mice were monitored for the development of disease symptoms twice daily for 14 days. Moribund animals meeting predefined endpoint criteria were humanely euthanized by CO₂ asphyxiation and scored as succumbing to infection 12 h later. For optical imaging, mice were anesthetized with isoflurane and imaged using the IVIS Spectrum imaging system (Caliper Life Sciences, Hopkinton, MA). Average radiance (photons/s/cm²) was calculated for ear and lymph nodes as previously described (Sun et al, 2012).

Statistics

All in vitro experiments were repeated three times and data are shown as the means ± standard deviations (SDs) from the three independent experiments, unless otherwise noted. Animal experiments were repeated twice to confirm reproducibility and the data is represented as the combination of the two independent experiments. P‐values were calculated using Student's t‐test, one‐way analysis of variance (ANOVA), or two‐way ANOVA with appropriate posthoc testing as indicated. All statistics were completed using GraphPad Prism software.

Author contributions

Sarah L Price: Conceptualization; data curation; formal analysis; funding acquisition; validation; investigation; writing – original draft; writing – review and editing. Derek Thibault: Conceptualization; data curation; formal analysis; validation; investigation. Taylor M Garrison: Data curation; formal analysis. Amanda Brady: Data curation. Haixun Guo: Data curation; formal analysis; writing – review and editing. Thomas E Kehl‐Fie: Conceptualization; resources; writing – review and editing. Sylvie Garneau‐Tsodikova: Conceptualization; writing – review and editing. Robert D Perry: Conceptualization; writing – review and editing. Tim van Opijnen: Conceptualization; resources; supervision; funding acquisition; methodology; writing – review and editing. Matthew B Lawrenz: Conceptualization; resources; formal analysis; supervision; funding acquisition; writing – original draft; writing – review and editing.

Disclosure and competing interests statement

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Dataset EV1

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 5

EMBO reports (2023) 24: e57369

Data availability

No large primary datasets have been generated and deposited.