A Serendipitous Mutation Reveals the Severe Virulence Defect of a Klebsiella pneumoniae fepB Mutant

In addition to having a reputation as the causative agent of several types of hospital-acquired infections, Klebsiella pneumoniae has gained widespread attention as a pathogen with a propensity for acquiring antibiotic resistance. It is capable of causing a range of infections, including urinary tract infections, pneumonia, and sepsis. Because of the rapid emergence of carbapenem resistance among Klebsiella strains, there is a dire need for a better understanding of virulence mechanisms and identification of new drug targets. Here, we identify the periplasmic transporter FepB as one such potential target.

and as a result, bacteria produce their own chelators to scavenge iron. Iron acquisition is an essential component of most bacterial pathogens, as iron is required for cellular and metabolic activities (18). Siderophores are small secreted molecules with a high affinity for ferric iron; these are classified on the basis of the chemical nature of the Fe 3ϩ coordination (19). The catecholate-type siderophore enterobactin is produced by most K. pneumoniae strains (20,21). However, community-acquired isolates and those that cause invasive disease typically encode additional siderophore systems (salmochelin, yersiniabactin, aerobactin) (22). Salmochelin is a C-glucosylated enterobactin produced by some isolates of Salmonella, Escherichia coli, and Klebsiella, and its synthesis is dependent on enterobactin. Mutants unable to produce enterobactin are also unable to produce salmochelin (23,24). The iroA locus encodes enzymes necessary to modify enterobactin, as well as proteins required for salmochelin transport (25). The yersiniabactin locus is found in many invasive K. pneumoniae isolates and encodes a phenolate-type siderophore that was first identified as part of a pathogenicity island in Yersinia (26). Interestingly, in a genome-wide association study of a broad range of K. pneumoniae isolates, yersiniabactin was found to be the most prevalent virulenceassociated locus and was found to be a predictor of infection versus carriage (22).
Aerobactin is yet another siderophore produced by a smaller fraction of K. pneumoniae strains than either enterobactin or yersiniabactin (22). Although aerobactin has a lower affinity for Fe 3ϩ than enterobactin or yersiniabactin, it is frequently produced by isolates from pyogenic liver abscesses (27).
To date, the identified virulence factors of K. pneumoniae primarily include capsule, lipopolysaccharide (LPS), fimbriae, and siderophores, and these factors also have been identified as virulence factors in the strain used for the studies presented here (4,(28)(29)(30)(31)(32)(33)(34). Several high-throughput studies have been done with mouse models to identify additional bacterial virulence factors (34)(35)(36)(37)(38)(39)(40). Two of these screens were signature-tagged mutagenesis (STM) screens for factors affecting gastrointestinal colonization and/or infection of the urinary tract (36,37). These studies identified adhesins, LPS, and capsule. Another screen for gain of function when Klebsiella genes were expressed in E. coli identified a response regulator, AcrA, and LPS (40). A screen for genes expressed in vivo during septicemia identified genes involved in the use of siderophores (aerobactin and enterobactin) (39), and an STM screen in a model of liver abscess formation identified adhesins and regulators (38). Two of these studies focused on the identification of bacterial genes needed for survival in the lung; one approach used STM, and the other used transposon insertion site sequencing (34,35). These screens identified capsule, LPS, siderophores, and transcriptional regulators. All of these screens also identified genes predicted to contribute generally to growth, as well as genes of unknown function.
Overall, there has been a lack of overlap in identified genes among the different screens conducted with lung, urinary tract, liver infection, and gastrointestinal colonization models. This may be due to the fact that none of the screens were saturating, or it could be indicative of mechanisms that compensate for the loss of individual genes. These findings are further complicated by the use of different infection models and different pathogen and host strain backgrounds. While typically focused on the goal of identifying previously unknown bacterial factors contributing to disease, these screens primarily identified known virulence factors of K. pneumoniae, as well as metabolic functions generally contributing to growth.
We previously conducted an STM screen of K. pneumoniae in an intranasal model of pneumonia to identify virulence genes (34). From this screen, yersiniabactin was identified as important for the abilities of our strain to colonize the lungs and to cause disseminated infection (33). In addition, a number of mutants with insertions in or near ramA were identified (34). RamA has been implicated in virulence and multidrug resistance in other pathogenic bacteria, and mutations in ramA have been associated with fluoroquinolone resistance in K. pneumoniae (41)(42)(43)(44). Furthermore, a recent study reported that overexpression of RamA affects virulence and results in modified LPS (45). Thus, we sought to determine if RamA is a virulence determinant for a highly virulent K. pneumoniae strain. These studies found no role for ramA or nearby genes for virulence in a pneumonia model of infection. However, a serendipitous secondary mutation was identified, and further analysis of this mutation indicates that FepB, a periplasmic protein required for transport of enterobactin and salmochelin, is essential for virulence. Surprisingly, there were interesting differences in virulence between enterobactin synthesis mutants and the ΔfepB mutant.

RESULTS
The smr mutant is severely attenuated in a mouse model of pneumonia. A previous screen of strain KPPR1 transposon mutants identified genes required for colonization and survival in the lungs of infected mice (34). Thirteen mutants containing disruptions within ramA or an adjacent gene, orf82, failed to be recovered from the lungs and spleens of infected mice. RamA is a transcriptional regulator linked to Salmonella survival in RAW 264.7 macrophages and virulence in BALB/c ByJ mice (41,42). This led us to hypothesize that the ramA locus is important for the ability of K. pneumoniae to infect the lungs. To test this, we constructed the smr (spontaneous multidrug resistance) mutant, where ramA and the two flanking genes (orf82 and romA) were targeted for deletion, and tested this strain in a mouse model of pneumonia (Fig. 1). The smr mutant caused slightly lower bacterial burdens at 24 h postinoculation (hpi) than KPPR1 (wild type [WT]). At 72 hpi, nearly 5 logs fewer CFU were recovered from mice infected with the smr mutant than from WT-infected mice. The spleens of mice infected with the WT strain had nearly 10 7 CFU/g of tissue, while the smr mutant was rarely detectable in the spleen at 72 hpi, reflecting a dissemination or systemic survival defect. Together, these data indicate that the smr mutant is essentially avirulent in this infection model.
Deletions of individual genes in the targeted smr locus do not recapitulate the phenotype of the smr mutant. To identify the gene(s) responsible for the phenotype of the smr mutant, we made in-frame deletions of each of the three genes (ΔramA, ΔromA, and Δorf82) in the smr locus and tested them in our pneumonia model ( Fig. 2A). The phenotype of all three mutant strains resembled that of the WT, suggesting that the loss of a single gene was not sufficient to affect virulence (Fig. 2B). We concluded that neither ramA, orf82, nor romA, individually contributed to virulence in this model or was responsible for the phenotype of the smr mutant.
In examining the region more closely, we noted that an RND (resistance-nodulationdivision superfamily) efflux pump system was encoded just upstream of orf82 and that the smr deletion could have impacted the promoter driving the expression of this locus ( Fig. 2A). RND efflux systems have been shown to play roles ranging from resistance to human antimicrobial peptides in Pseudomonas to flagellar motility in Burkholderia (46).  Thus, we constructed two additional mutants, one with the rnd genes and the other with orf82, ramA, and romA deleted but with the putative rnd promoter intact (Δrnd and Δorf82 ramA romA). The Δrnd mutant colonized mice as efficiently as the WT strain (Fig. 2B). Intriguingly, the second mutant lacking the same three genes as the smr mutant (Δorf82 ramA romA) also had no virulence defect. Sequencing of the smr mutant reveals a large deletion. As targeted genetic mutations in the smr locus failed to recapitulate the smr phenotype, we hypothesized that the smr mutant contained a secondary mutation. Whole-genome sequencing revealed that the deletion in the smr mutant was larger than intended. Instead of the targeted deletion of orf82, ramA, and romA, a single segment of 87,290 bp spanning 78 putative open reading frames was deleted.
A component of the enterobactin transport system contributes to virulence. To identify the factor(s) responsible for the virulence defect of the smr mutant, we constructed three mutants (Δsmr_A, Δsmr_B, and Δsmr_C) each with a deletion of approximately one-third of the genes deleted in the smr mutant (Fig. 3A). The putative orf genes in each mutant are listed in Table 1. In our pneumonia model at 24 and 72 hpi, both Δsmr_A and Δsmr_B mutant-infected mice had bacterial burdens comparable to those of mice infected with the WT (Fig. 3B). However, the mice infected with Δsmr_C mutant had Ͼ1 log fewer CFU/g at 24 hpi and nearly 6 logs fewer CFU/g at 72 hpi than mice infected with the WT. Thus, the Δsmr_C mutant recapitulated the phenotype of the smr mutant, whereas the Δsmr_A and Δsmr_B mutants behaved like the WT strain.
Located within the region deleted in the Δsmr_C mutant are genes necessary for the synthesis, export, and import of the siderophore enterobactin. We therefore hypothesized that a component of the enterobactin transport system was responsible for the virulence defect of the smr mutant. We did not believe that the siderophore itself was responsible, as an ΔentB mutant, which is unable to synthesize enterobactin and salmochelin, is only modestly attenuated in this mouse pneumonia model (33). The enterobactin receptor FepA also was not implicated, as FepA is encoded within the region deleted in the Δsmr_B mutant. inverted triangles) mutant. At 24 or 72 hpi, mice were sacrificed and their lungs and spleens were homogenized and plated for bacterial enumeration. Each symbol represents one mouse. The dotted line indicates the limit of detection, and symbols on the dotted line indicate that CFU counts were below the limit of detection. These data were compiled from several independent experiments. Mann-Whitney tests were performed for statistical analysis. ***, P Ͻ 0.001; ****, P Ͻ 0.0001.
Siderophore transport involves several membrane proteins. For enterobactin, EntS and TolC are required for export, whereas FepA, FepDGC, and Fes are required for import. In addition, the periplasmic protein FepB is required for the import of both enterobactin and salmochelin. Because previous studies had implicated siderophore transport components in virulence (47), we targeted specific components of the enterobactin siderophore transport system and tested loss-of-function (ΔentS, Δfes, ΔfepB, and fepD::pKAS46) mutants in our pneumonia model (Fig. 4A). We included a different enterobactin synthesis (ΔybdB2 entABEC [referred to as Δentsyn]) mutant to confirm our previous findings obtained with the ΔentB mutant (33). We found that only the ΔfepB mutant recapitulated the phenotype of the smr mutant, as demonstrated by the attenuation in the lungs and the lack of dissemination at 24 and 72 hpi (Fig. 4B). Consistent with previously studies, neither the Δentsyn mutant ( Fig. 4B) nor the ΔentB mutant ( Fig. 5) recapitulated the smr phenotype (33). In addition, loss of fepB did not affect the expression of the yersiniabactin system ( Fig. 6), consistent with results previously obtained with an enterobactin synthesis mutant (33). Thus, the periplasmic transport protein FepB contributes to virulence in a manner distinct from that of enterobactin and salmochelin uptake alone.
A variety of different approaches were used to complement the ΔfepB mutant, but all were unsuccessful. Plasmid-based approaches failed to complement growth under iron-depleted conditions, despite the constitutive expression of fepB (data not shown). We also attempted to repair the deletion, but this strain could not be obtained, for reasons we do not understand. Problems with fepB complementation are not unprecedented and were also reported for a Salmonella fepB mutant (47). To ensure that the observed phenotype of the ΔfepB mutant was not a consequence of secondary mutations, a second fepB mutant (fepB2) was constructed and found to recapitulate the virulence and growth phenotypes of the original fepB mutant (Fig. 5). Additionally, we sequenced across the deletion junction of both of the ΔfepB mutants and obtained the expected sequence, suggesting that a larger deletion of the region surrounding fepB had not occurred (data not shown). Expression of the genes adjacent to fepB, entC and entS, was assessed by quantitative reverse transcription-PCR. Expression of entC and entS was not detected in the ΔfepB mutant but was in the WT (data not shown). EntC and EntS may be needed for growth under low-iron conditions, and their lack of expression provides a possible explanation for failed complementation in trans. How-   4A) (33). Thus, we conclude that deletion of fepB results in a phenotype distinct from that of other enterobactin system mutants.
A fepB mutant resembles a ⌬entB ⌬ybtS double mutant. We previously showed that a ΔentB ⌬ybtS mutant that is deficient in all siderophore production was severely attenuated (33). In comparing the defect of the ΔfepB mutant strain to those of other siderophore mutants, we noticed that the phenotype of the ΔfepB mutant was similar to that of the ΔentB ⌬ybtS mutant. Because the attenuation of the ΔfepB mutant was much greater than that of the ΔentB mutant, we hypothesized that the role of FepB is not limited to enterobactin import and that it might be involved in an additional iron acquisition system. To gain a better understanding of the relationship between the phenotypes of these mutants, we tested the ΔfepB mutant together with the ΔentB ⌬ybtS mutant to determine if its virulence defect resembles that of a ΔentB ⌬ybtS mutant in vivo and included a ΔentB mutant as a control (Fig. 5). The ΔfepB and ΔentB ⌬ybtS mutants had similar attenuation levels, which were more severe than that of the Proofreading thioesterase in enterobactin biosynthesis, YbdB2 VK055_1920 Carbon starvation CstA family protein VK055_1919 Helix-turn-helix family protein VK055_1918 Hypothetical protein VK055_1917 Plasmid stabilization system family protein VK055_1916 Short-chain dehydrogenase family protein VK055_1915 Iron-containing alcohol dehydrogenase family protein VK055_1914 ABC transporter family protein VK055_1913 Branched-chain amino acid transport system/permease component family protein VK055_1912 Periplasmic binding and sugar binding domain of LacI family protein VK055_1911 LVIVD repeat family protein a HAD, haloacid dehalogenase. b TPP, thiamine pyrophosphate. To address the role of FepB in iron uptake and to determine if the virulence defect could be due to reduced iron acquisition, we used an in vitro growth model. The ΔfepB, ΔentB, and ΔentB ⌬ybtS mutants were grown in defined medium with or without the iron-chelating agent 2,2=-dipyridyl (DP). All of the strains had similar growth rates in the absence of DP, indicating that the mutants grow normally when iron levels are sufficient (Fig. 7A). However, in the presence of DP, the growth of the ΔfepB and ΔentB ⌬ybtS mutants was severely restricted (Fig. 7B). Interestingly, the growth of the ΔentB mutant was restricted compared to that of the WT strain, but the triple siderophore (ΔentB ⌬ybtS) mutant and the ΔfepB mutant grew even more slowly than the ΔentB mutant. These data suggest that FepB contributes to growth in an irondependent manner that is distinct from its known role in enterobactin and salmochelin uptake.
Yersiniabactin import is unaffected in a ⌬fepB mutant. The ΔfepB mutant had a stronger phenotype than an enterobactin/salmochelin synthesis mutant, and it resembled that of a triple siderophore mutant in both virulence and growth under iron limitation. Yersiniabactin is the only known siderophore produced by the ΔentB mutant but not the ΔentB ⌬ybtS mutant. Thus, we wanted to assess if the ΔfepB mutant is defective in yersiniabactin uptake. To do this, we performed a cross-feeding experiment to determine if the growth defect of the ΔfepB mutant under iron-limited conditions could be restored in the presence of yersiniabactin by coculturing the ΔfepB mutant with a yersiniabactin-producing strain. We predicted that if FepB is required for yersiniabactin import, a feeder strain producing yersiniabactin would be unable to restore the growth of the ΔfepB mutant. In this assay, test strains were spread onto M9 medium supplemented with 0.4% glucose and 0.2% Casamino acids (M9-CAA) agar containing DP and feeder strains were then spotted onto the surface of the plates. The WT and ΔentB, ΔentB ⌬ybtS, and ΔfepB mutant strains were used as test strains, and the WT and the ΔentB (capable of producing yersiniabactin) and ΔybtS (does not produce yersiniabactin) mutants were used as feeder strains. As expected, the ΔybtS mutant was not able to complement the growth defect of the ΔfepB mutant, as the ΔfepB mutant should not be able to use the enterobactin produced by this strain (Fig. 8A). The WT and the ΔybtS mutant were able to complement the growth of the ΔentB ⌬ybtS mutant, as expected (Fig. 8B). Importantly, the ΔentB mutant and the WT were able to restore the growth of the ΔentB mutant (as expected), as well as the ΔfepB mutant. This finding suggests that yersiniabactin can still be imported by a ΔfepB mutant.
To determine if the complementation of the ΔfepB mutant's growth defect by a yersiniabactin-producing strain in the cross-feeding experiment was due to yersiniabactin production rather than the production of other secreted bacterial products, we performed a similar experiment by spotting purified apo-yersiniabactin instead of feeder strains. As described above, test strains (WT strain and ΔfepB and ΔentB ⌬ybtS mutants) were spread onto M9-CAA agar containing DP. Various concentrations of apo-yersiniabactin were applied to paper discs that were placed on the agar plate to test for growth restoration and thus the ability to utilize yersiniabactin (Fig. 8C). The WT strain was able to grow even without yersiniabactin supplementation. The ΔentB ⌬ybtS and ΔfepB mutants did not grow around the vehicle control (distilled H 2 O [dH 2 O]) disc. However, upon the addition of yersiniabactin, the growth defect of the ΔentB ⌬ybtS mutant was restored in a concentration-dependent manner; this is an expected result because this strain is still able to import exogenous yersiniabactin. Addition of apoyersiniabactin also restored the growth of the ΔfepB mutant (Fig. 8C). Together, these data suggest that FepB is not required for yersiniabactin import in vitro and that the virulence defect of the ΔfepB mutant is due to a mechanism unrelated to yersiniabactin import. Capsule production is not responsible for the ⌬fepB mutant's phenotype. Capsule is considered a primary virulence factor of K. pneumoniae (reviewed in reference 4). Therefore, to test if there was a change in capsule production that could contribute to the ΔfepB mutant's phenotype, we measured its uronic acid content. When the ΔfepB mutant and the WT strain were grown in Luria-Bertani (LB) medium at 37°C, the same conditions used for the inoculum used in mouse experiments, there was no difference in capsule production (Fig. 9A). Similarly, when mucoviscosity was measured (another assay for capsule phenotypes), we saw no measureable difference between the WT and the ΔfepB mutant (Fig. 9B).
Because iron levels can affect K. pneumoniae capsule production (48), we decided to test if capsule production is altered in the ΔfepB mutant under low-iron conditions. All four siderophore system (ΔfepB, ΔentB, ΔybtS, and ΔentB ⌬ybtS) mutants had a modest, nonsignificant reduction in capsule production (Fig. 9C). The mucoviscosity of the siderophore system mutants was also lower than that of the WT (Fig. 9D). Importantly, there was no difference between the capsule production levels of the ΔfepB and ΔentB mutants. How FepB affects virulence is not clear, but it does not appear to be related  to the amount of capsule produced ( Fig. 9A and C)) or the mucoviscosity of the capsule ( Fig. 9B and D), as the uronic acid content and sedimentation of the ΔfepB mutant were comparable to those of the enterobactin synthesis mutant, which is only modestly attenuated.

DISCUSSION
The repertoire of confirmed K. pneumoniae virulence factors has changed little during the past 2 decades (2, 4). Although a number of large screens for K. pneumoniae virulence determinants have been performed (34)(35)(36)(37)(38)(39)(40), unfortunately, there have been few follow-up analyses of the results of these screens. In a screen of signature-tagged mutants in a pneumonia model of infection, we identified a locus that included ramA as potentially important for virulence (34), and a recent study suggested that overexpression of ramA affects virulence and leads to LPS modifications (45). In this study, we constructed a mutant (smr) with this locus deleted and found that it was cleared from the lungs following intranasal inoculation and that it was unable to spread systemically. Why deletion of ramA or the surrounding genes did not result in a virulence defect in the lungs and/or spleen when 11 insertions in this region were identified in the STM screen remains a mystery (34). One possibility is that in the STM screen, each insertion mutant was screened essentially in competition with 95 other mutants, most of which behave like the WT strain. Therefore, a ramA mutant may have a competitive disadvantage when at a ratio of~1:100 with the WT but will not exhibit a defect when inoculated on its own. RamA has been implicated in the regulation of pathways important for multidrug resistance (43,44), and thus, it may still be important in the context of antibiotic treatment or in a strain background that is not hypervirulent. Subsequent analysis of the smr mutant indicated that the virulence defect was due not to deletion of the ramA locus but rather to the deletion of fepB, a gene encoding Overnight cultures of the WT strain, the ΔfepB mutant, and a capsule-deficient strain (cpsB::Tn5Kn2) were subcultured to an OD 600 of 0.2 and grown in LB medium for 6 h, and total capsule production was measured with the uronic acid assay (A) and the low-speed centrifugation assay to measure mucoviscosity (B). These data are from strains grown in triplicate in an individual experiment. One-way analysis of variance, followed by Dunnett's multiplecomparison test, was performed for statistical analysis. ***, P Ͻ 0.001; ****, P Ͻ 0.0001; ns, not significant.
FepB and Klebsiella Virulence a protein required for enterobactin and salmochelin import (49)(50)(51). The fepB mutant had a more severe growth defect in iron-limited medium and a more severe in vivo defect than an enterobactin synthesis (ΔentB) mutant; the ΔentB mutant would also be deficient in salmochelin production. The contributions of the siderophores enterobactin, salmochelin, and yersiniabactin to Klebsiella virulence have been examined previously, and individually, they were found to contribute only minimally to infection (32,33,52). The data presented here reveal that while enterobactin/salmochelin may be dispensable for the virulence of a strain also able to produce yersiniabactin in a K. pneumoniae lung infection model, the enterobactin/salmochelin importer FepB is necessary to establish infection. Furthermore, under both in vitro and in vivo conditions, the ΔfepB mutant resembles a ΔentB ⌬ybtS mutant, which is unable to produce any of the three siderophores encoded by this strain (enterobactin, salmochelin, and yersiniabactin). Together, these observations suggest that FepB contributes to virulence and growth under iron limitation in an unanticipated way. Siderophores are synthesized in the cytoplasm and require machinery for export and subsequent import following iron sequestration. Enterobactin is synthesized by EntABCDEF and is exported to the periplasm via the inner membrane protein EntS and subsequently through the outer membrane via the membrane channel protein TolC (53). Once bound to ferric iron, enterobactin (enterobactin-Fe 3ϩ ) binds the outer membrane siderophore receptor FepA and is translocated into the periplasm by a TonB-dependent mechanism. In the periplasm, enterobactin-Fe 3ϩ then binds the periplasmic chaperone FepB and is shuttled to the inner membrane, where it interacts with the inner membrane transport complex FepDGC and is ultimately released into the cytoplasm (50,54,55). Salmochelin utilizes a similar export apparatus but is imported via the bacterial outer membrane receptor IroN, and then FepB shuttles it to FepDGC (56). Export and yersiniabactin import appear to be similar, although several steps in yersiniabactin transport remain to be elucidated (53). Specifically, no periplasmic protein (FepB equivalent) has been identified in the yersiniabactin import system. Because of the similarities in the phenotypes of the ΔfepB and triple siderophore mutants and because no FepB equivalent has been identified in the yersiniabactin import system, we initially hypothesized that FepB may be involved in yersiniabactin import. However, our results show that a ΔfepB mutant can still utilize yersiniabactin for growth in vitro, and thus, the role of FepB in growth under iron limitation and virulence remains unclear. A recent crystal structure of FepB indicates that it can form a trimer (57) and thus possibly could coordinate a target other than enterobactin-Fe 3ϩ , but this has yet to be demonstrated.
A contribution of the periplasmic enterobactin transporter FepB to pathogenesis also was observed in Salmonella enterica (47). Salmonella produces both enterobactin and salmochelin, and both siderophores require FepB for import (25). However, Nagy et al. found that a fepB mutant had lower colonization levels in mice than a fepA-iroN double mutant (encoding the outer membrane receptors for enterobactin and salmochelin) in a gastric model of infection (47,58). This is comparable to our results obtained with K. pneumoniae and suggests that the role of FepB in virulence extends beyond siderophore transport. The fact that this phenomenon has been reported in two Gram-negative pathogens hints that this may be a conserved mechanism in other bacterial species. One possible explanation for this observation is that in a ΔfepB mutant, enterobactin is not recycled properly and accumulates extracellularly and perhaps this is detrimental to the bacteria, given that enterobactin can enhance copper toxicity (59). However, in this scenario, the Δsmr_C mutant (which is a ΔentB ΔfepB double mutant and has other genes [listed in Table 1] deleted) should not have this phenotype, as it would be unable to produce enterobactin. However, the data presented here suggest that this is not the case, as the Δsmr_C mutant has a virulence defect comparable to that of a ΔfepB mutant.
Interestingly, recent studies have noted that the complement of siderophore systems produced by an individual strain of K. pneumoniae has a significant impact on its ability to colonize versus its ability to cause an infection or its ability to cause invasive disease associated with the hypervirulence phenotype (22). In an analysis of a broad sampling of over 300 strains, only 33% of an individual strain's genome is part of the core Klebsiella genome, and the remaining 67% is composed of "accessory" genes that vary significantly from strain to strain (22). Until recently, the gene profiles necessary to cause the different types of infections associated with K. pneumoniae were not clear. However, recent bioinformatics analyses of large strain collections, combined with information on the type of infection, have revealed that some specific gene profiles are associated with colonization versus infection versus invasive disease. For example, the presence of rmpA (a regulator of capsule), as well as the genes required for the production and use of the siderophores aerobactin, salmochelin, and yersiniabactin, was highly associated with strains isolated from infections versus carriage alone (22). Interestingly, an additional five loci were associated with invasive infections (versus noninvasive infections or carriage), including fepB. This is consistent with the requirement we observed for fepB to cause disseminated infection in mice and what has been observed in Salmonella (47).
With antibiotic resistance on the rise, the development of new therapeutics to combat infection by multidrug-resistant bacteria is an urgent need (60). Siderophore systems present an attractive target for drug development because of the conservation of these systems among Gram-negative pathogens (61). Immunization with the yersiniabactin receptor FyuA or the siderophores themselves (yersiniabactin and aerobactin) was protective when tested in a murine model of E. coli urinary tract infection (62)(63)(64). FepB may be an especially attractive target to consider for drug development, as it is required for disseminated infections and is found in a wide variety of bacteria. In addition to being potential targets for drug development, siderophores represent an attractive system to exploit as a drug delivery mechanism to overcome the permeability barrier of the outer membrane. In essence, the siderophore can be used as a "Trojan horse" to target a siderophore-drug conjugate to the siderophore-iron transport systems (61). This would allow the delivery of drugs to the periplasm and potentially to the cytoplasm. From the work presented here and with Salmonella, one such periplasmic target could be FepB itself. Drug-siderophore conjugates have been developed, and a catechol-cephalosporin conjugate, cefiderocol (S-649266), was found to have lower MIC 90 s than the antibiotics cefepime, piperacillin-tazobactam, and meropenem when tested against several Gram-negative bacteria, including multidrug-and carbapenemresistant strains (65)(66)(67). Cefiderocol displayed antibacterial properties when tested in vivo and is currently being tested in a phase 3 clinical trial against carbapenem-resistant Gram-negative infections in humans (66,68). Thus, investigations probing the mechanisms of siderophore transport can provide the basis for promising new therapeutics.

MATERIALS AND METHODS
Ethics statement. Mouse experiments were conducted in accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health (69). All animal studies were approved by the Institutional Animal Care and Use Committee at the University of North Carolina (UNC) at Chapel Hill (protocols 11-127 and 14-110). All efforts were made to minimize suffering. Animals were monitored daily following inoculation and were euthanized upon exhibiting signs of morbidity.
Bacterial strains and culture conditions. The bacterial strains and plasmids used in this study are described in Table 2. The WT parental strains are KPPR1, a Rif r derivative of ATCC 43816 (34), and KPPR1S, a Str r derivative of KPPR1; they have identical growth characteristics in vitro and in vivo. K. pneumoniae strains were grown aerobically in LB medium or M9-CAA overnight at 37°C. Where indicated, 100 or 200 M DP (Sigma-Aldrich, St. Louis, MO) was added to M9 or LB medium, respectively, to deplete the available iron. Antibiotics were added to the medium as appropriate at the following concentrations: kanamycin, 50 g/ml (Kan 50 ); rifampin, 30 g/ml (Rif 30 ); streptomycin, 500 g/ml (Strep 500 ). Bacterial growth was monitored by measuring the optical density at 600 nm (OD 600 ).
Construction of bacterial mutants. Mutations in KPPR1S (ΔramA Δorf82, Δorf82ΔramAΔromA, ΔentS, ΔybdB2 entABEC [referred to as Δentsyn], Δfes, ΔfepB, Δsmr_A Δsmr_B, and Δsmr_C) were generated by allelic exchange by using pKAS46, a suicide vector that allows the use of streptomycin for counterselection (70,71). Sequences up-and downstream (~500 bp each) were generated by PCR with the primer sets indicated in Table 3, cloned into pKAS46, and confirmed by sequence analysis. Overnight cultures of KPPR1S and E. coli S17-1 pir (72) carrying a derivative of pKAS46 were mixed, collected by centrifugation, plated on LB agar (no antibiotics), and grown overnight at 37°C. Transconjugants were selected by plating on LB agar with Rif 30 and Kan 50 . Several Rif r Kan r colonies were grown for 5 to 6 h in LB medium (no antibiotics) and then plated on LB agar with Strep 500 to select for transconjugants that had excised the plasmid. Kan s clones were screened by PCR to verify the loss of the targeted gene(s). An insertional disruption of the fepDCG operon was constructed in KPPR1S (fepD::pKAS46) by plasmid integration into the fepD gene. A DNA fragment generated by PCR with primers MP313 and MP314 (Table 3) was cloned into pKAS46. The resulting plasmid, pKAS46fepD::kan, was conjugated into KPPR1S as described above. Colonies with integration of the plasmid on the chromosome were identified by plating on LB agar with Rif 30 and Kan 50 .
Isogenic mutants of KPPR1 (ΔromA and Δsmr) were generated by allelic exchange with pKO3 as previously described (73). pKO3 is a vector that allows the use of sucrose as a positive selection for the loss of the vector. DNA fragments were amplified by PCR with the primer sets indicated in Table 3 and cloned into pKO3, generating plasmids pKO3ΔromA and pKO3Δsmr.
Whole-genome sequencing of the smr mutant. Total DNA from the smr mutant (VK82) was isolated with a genomic DNA purification kit (Qiagen), and the sample was submitted to the UNC High-Throughput Sequencing Facility for sequencing. An Illumina HiSeq 2000 instrument generated 2 ϫ 75-bp paired-end reads. A mapped genome assembly was produced with the "Map Reads to Reference" tool in CLC Genomics Workbench v7.5.1 by using the published KPPR1 genome as the template (74). The smr mutant and KPPR1 parent strain genomes were then compared with the "Basic Variant Detection" tool in CLC Genomics Workbench to identify mutations in the smr strain. Mutations were visualized by aligning these genomes with Mauve (75). ybtA promoter in pPROBE 33 Murine model of pneumonia. Five-to 8-week-old, female C57BL/6 mice (Jackson Laboratories) were anesthetized by intraperitoneal injection with 200 l of a mixture of ketamine (6.66 mg/kg) and xylazine (10.6 mg/kg). Overnight bacterial cultures were diluted in phosphate-buffered saline (PBS), and 20 l was inoculated intranasally (i.n.) in two 10-l aliquots for a total of~2 ϫ 10 4 CFU/mouse as previously described (34). At 24, 48, or 72 hpi, mice were euthanized by a lethal injection of 200 l of sodium pentobarbital (150 mg/kg). Organs were removed, homogenized in PBS, serially diluted, and plated to quantify the number of CFU/g of tissue.
Mucoviscosity assay. Mucoviscosity was determined as previously described (35,76). Briefly, overnight cultures were grown in LB medium, subcultured to an OD 600 of 0.2 in fresh medium, and grown at 37°C. After 6 h, cultures were normalized to 1.0 U of OD/ml and centrifuged for 5 min at 1,000 ϫ g and the OD 600 of the supernatant was measured.

Primer
Sequence a (5= to 3=) Description  MP66  TGACTAGATATCGCTGATTACCGAAGCGGACTG  5= flank forward ΔramA  MP67  TGCATATCTAGAGGAAATCGTCATATGCTCTCT  5= flank reverse ΔramA  MP68  TGCATATCTAGACACTGAGGCGCGCCTCTCCTG  3= flank forward ΔramA  MP69  TCGATAGCGGCCGCCGACTGGCGCTGTACATCGCG  3= flank reverse ΔramA  MP114  TGACTAGATATCTCGCCCGAGGGCGTCGTAAAC  5= flank forward Δorf82  MP71  TGCATATCTAGACTCGAGCGGTAAACCAGGAGA  5= flank reverse Δorf82  MP72  TGCATATCTAGACAGTGGATGTTTCATGTCATG  3= flank forward Δorf82  MP115  TCGATAGCGGCCGCGGGATGAACCGTATCAACGGC  3= flank reverse Δorf82  MP124  TGACTAGATATCCGATTTTGCCTGCTATGCGCA  5= flank forward Δrnd  MP125  TGCATATCTAGACATCGGCGGGGGTAAGCGCGG  5= flank reverse Δrnd  MP126  TGCATATCTAGAGTTCACCCGGTCGCCCAGCGG  3= flank  Extraction and quantification of capsule. Uronic acid was extracted and quantified as previously described (28). Briefly, overnight cultures were grown in LB medium, subcultured to an OD 600 of 0.2 in fresh medium, and grown at 37°C. After 6 h, 500 l of culture was added to 100 l of 1% Zwittergent-100 mM citric acid and incubated at 50°C for 20 min. Cells were pelleted, and 300 l of the supernatant was added to 1.2 ml of absolute ethanol, incubated at 4°C for 20 min, and centrifuged for 5 min at maximum speed. The pellet was resuspended in 200 l of dH 2 O, added to 1.2 ml of 12.5 mM sodium tetraborate in sulfuric acid, and incubated for 5 min at 100°C. A 20-l volume of 0.15% 3-phenylphenol was added, and the absorbance at 520 nm was measured. The glucuronic acid content was determined from a standard curve of glucuronic acid (Sigma-Aldrich, St. Louis, MO) and expressed in micromoles per OD unit.
Measurement of promoter activity. Expression of the yersiniabactin-encoding locus was assessed in vitro with a transcriptional gfp reporter containing the sequence 500 bp upstream of the ybtA promoter cloned into pPROBE (33,77). The bacteria were grown overnight at 37°C in LB medium, subcultured to an OD 600 of 0.2, and grown for 6 h with or without 200 M DP. All strains were assayed in triplicate. Fluorescence was detected with a Synergy HT microplate reader (BioTek Instruments, Winooski, VT) and measured in relative fluorescence units per OD 600 unit.
In vitro growth curves. To monitor bacterial growth, bacterial strains were grown overnight in M9-CAA at 37°C, subcultured to an OD 600 of 0.05 in fresh medium in 250-ml flasks, and grown with aeration for 6 h at 37°C. OD 600 readings were recorded at the intervals indicated. Medium was supplemented with 100 M DP to examine bacterial growth under iron-limiting conditions.
To determine if secreted siderophores could restore the growth of siderophore mutants in iron-depleted medium, a cross-feeding assay was performed as previously described, with minor modifications (78). Bacteria were grown overnight at 37°C in M9-CAA. Approximately 1 ϫ 10 5 CFU of each test strain was spread onto M9-CAA agar plates containing 100 M DP. Feeder strains were then spotted (2.5 l of overnight culture) onto the agar, and the plates were incubated at 37°C overnight.
To determine if purified yersiniabactin could restore the growth of siderophore mutants in irondepleted medium, test strains were spread on M9-CAA agar as described above. Iron-free yersiniabactin (EMC Microcollections, Germany) was resuspended in ethanol, and 10 l of either 1 mM or 100 M yersiniabactin (diluted in dH 2 O) was spotted onto filter disks on the plate to assess yersiniabactindependent growth complementation.