Iron Acquisition and Siderophore Release by Carbapenem-Resistant Sequence Type 258 Klebsiella pneumoniae

Carbapenem-resistant Enterobacteriaceae, including K. pneumoniae, are a major health care concern worldwide because they cause a wide range of infection and are resistant to all or nearly all antibiotics. To cause infection, these bacteria must acquire iron, and a major mechanism of acquiring iron is by secreting a molecule called enterobactin that strips iron from host proteins. However, a subset of carbapenem-resistant K. pneumoniae strains that lack a portion of the entS gene that is required for enterobactin secretion was recently discovered. To understand how these mutant strains obtain iron, we studied their transcriptional responses, bacterial growth, and enterobactin secretion under iron-limited conditions. We found that strains both with mutated and intact entS genes grow under iron-limiting conditions, secrete enterobactin, and utilize an alternate iron source, hemin, for growth. Our data indicate that carbapenem-resistant K. pneumoniae can use varied methods for iron uptake during infection.

C arbapenem-resistant Enterobacteriaceae (CRE) are resistant to all or nearly all antibiotics, are increasingly present worldwide, and are therefore designated an urgent public health concern (1). Infection with carbapenem-resistant (CR) Klebsiella pneumoniae is of particular concern due to its high mortality rate of approximately 41 to 50% for bloodstream infections (1-3). Klebsiella species cause a variety of infections, including pneumonia, urinary tract infection, and bacteremia, and represent the third most common cause of health care-acquired infection, further contributing to their relevance to health care settings (4). Despite epidemiological studies documenting the spread of infection and patient risk factors associated with infection, the bacterial pathogenesis of CR K. pneumoniae is currently not well understood (5)(6)(7).
To measure the genomic diversity of CR K. pneumoniae, a cohort of isolates was collected from hospitals within the midwestern United States for whole-genome sequencing (8,9). Strains isolated from infection and colonization were primarily the multilocus sequence type (MLST) ST258 (sequence type 258), but molecular genotyping based on core genome single nucleotide variants separated them into two distinct clades, ST258a and ST258b (8,9). ST258a isolates were significantly associated with the carbapenemase bla KPC-2 , whereas ST258b isolates were associated with the carbapenamase bla   (8). Significant epidemiological differences were present as well. Infection with ST258a isolates was more likely to result in an increased length of hospital stay than ST258b infection, and ST258b infection was more commonly acquired from a skilled nursing facility, whereas ST258a infections were more commonly community acquired (10). These differences in presentation and outcome suggested functional differences between these clades during infection. Indeed, whole-genome comparisons uncovered numerous polymorphisms within the capsular polysaccharide locus and extending into the mdtABC locus, as well as differences in plasmid contents (8,9). Intriguingly, a 347-bp deletion in the enterobactin (Ent) siderophore exporter gene entS (entSΔ 347 ) was found in a subset of ST258b isolates that was absent in ST258a isolates (9). The entS nucleotide sequence of ST258a isolates contains a 10-nucleotide repeat, and the deletion in ST258b isolates begins after the first instance and ends immediately after the second repeat (nucleotides 324 to 670; see Fig. S1A in the supplemental material). This mutated allele (entSΔ 347 ) contains a frameshift, altering the amino acid sequence beginning at position 109 and resulting in a premature stop codon truncating the protein at position 136 compared to the full-length 413-aminoacid sequence (Fig. S1B).
Ent is the prototypic catecholate siderophore and is secreted by Gram-negative bacteria, including K. pneumoniae, to acquire iron (11). Siderophores are major virulence factors in K. pneumoniae, and Ent has the highest known affinity for iron of any molecule (11). To counter the iron-sequestering effects of Ent, host cells secrete lipocalin 2 (Lcn2) that specifically binds to and sequesters Ent. Because entS encodes the only known Ent exporter, a deletion in entS is predicted to impair the ability of bacteria to secrete Ent for iron acquisition (12). Because K. pneumoniae ST258 isolates do not carry genes encoding additional siderophores, the inability to secrete Ent could be deleterious. In Escherichia coli, entS-null mutants release Ent breakdown products that have lower affinity for iron than that of intact Ent. The Ent esterase Fes cleaves cyclic Ent into monomers, dimers, and trimers of 2,3-dihydroxybenzoylserine (DHBS) (13,14). Whereas the affinity of the linearized trimer of DHBS is similar to that of Ent, the iron affinities of DHBS monomers and dimers are less than that of Ent (15). Despite their lower affinities, some of these breakdown products can act as siderophores (13,14). In extraintestinal pathogenic E. coli, an entS mutant is significantly attenuated in a systemic infection model (12). However, in K. pneumoniae, conservation of the entSΔ 347 allele within a subset of the ST258b clade suggests that compensatory iron acquisition or alternative Ent transport mechanisms may be present. A decrease in the export of Ent in favor of release of linear DHBS molecules and use of alternate compensatory iron acquisition pathways could be a strategy for some K. pneumoniae strains to evade the bacteriostatic effects of Lcn2 during infection.
To understand iron acquisition by CR K. pneumoniae, we examined the transcrip-tional and phenotypic responses of ST258 isolates to iron limitation induced by incubation in minimal medium and human serum. To study gene transcription in response to iron-replete and iron-depleted conditions, we performed transcriptome sequencing (RNA-Seq) to identify differential gene expression based on iron status and entS deletion. Additionally, we performed in vitro assays to characterize the ability of ST258a and ST258b strains to grow in iron-limiting medium and to secrete ironchelating molecules. We characterized the secretion of Ent by these strains using mass spectrometry. These results serve to increase the general understanding of CR K. pneumoniae iron acquisition and identify distinct iron starvation responses by ST258a and ST258b CR K. pneumoniae.

RESULTS
RNA-Seq analysis identifies differential gene expression in response to irondepleted media. The representative strains of K. pneumoniae utilized in this study are shown in Table 1. To examine the effect of the entSΔ 347 allele on the transcriptional response to iron-limited media, RNA-Seq was performed on the following strains: KPPR1, a wild-type (WT) strain secreting the siderophores Ent, yersiniabactin (Ybt), and salmochelin (Sal); UHKPC05, an ST258a strain with a full-length entS gene; NJST258_2, an ST258b strain with a full-length entS gene; UHKPC48 and VAKPC297, ST258b strains with the entSΔ 347 allele. Samples were collected for RNA-Seq 1 h after subculture into iron-limited minimal medium alone (M9/C) or medium supplemented with iron (M9/ CϩFe) or 5% human serum (M9/Cϩserum; see Fig. S2 in the supplemental material). To identify genes among all ST258 isolates that were significantly modulated under iron-depleted conditions (iron-depleted M9 minimal medium [M9/Chelex] versus ironreplete M9 medium [M9/ChelexϩFe]), the DESeq2 package in R was used to analyze count data obtained from RNA-Seq (Table 2). Negative fold changes represent genes that were significantly downregulated in response to iron depletion, whereas positive fold changes represent genes that were significantly upregulated in response to iron depletion. Genes that were significantly downregulated in response to iron-depleted media include the iron storage protein ferritin gene (KPNJ2_01976) and superoxide dismutase sodB (KPNJ2_02431) that uses iron as a cofactor. The downregulation of these iron-dependent genes during iron-depleted conditions serves as an internal validation. Genes that were significantly upregulated in response to iron-depleted media include Ent synthesis genes (reference gene loci KPNJ2_04011-04015), an ironsulfur cluster assembly locus (KPNJ2_02279-02282), a proposed TonB-dependent siderophore receptor fitA (Fig. 1A), a catecholate siderophore receptor fiu locus (Fig. 1B), and a hemin transport system gene (KPNJ2_01180-01184). To identify transcriptional differences associated with the entSΔ 347 allele, gene expression in entS ϩ and entSΔ 347 isolates were compared in the physiologically relevant iron-limited condition M9/ Chelexϩserum. Expression of entS was reduced in entSΔ 347 isolates ( Fig. 2A, 2.3 log 2 fold change and adjusted P value of 2.72EϪ6). In contrast, gene expression of the hemin transporter hmuS was significantly higher in entSΔ 347 isolates than in entS ϩ isolates (Fig. 2B, Ϫ1.71 log 2 fold change and adjusted P value of 7.14EϪ2).

Growth of K. pneumoniae ST258 strains in iron-replete and iron-depleted conditions.
To determine the ability of ST258 strains to grow under iron limitation, cultures were extensively diluted (1:100,000) into iron-depleted M9 minimal medium (M9/Chelex), and growth curves were constructed. Under these conditions, an E. coli entS-null mutant had a small but consistent growth defect (Fig. S3), suggesting that a functional defect in ST258 entSΔ 347 mutants may be detected. Because the extreme antibiotic resistance of ST258 eliminates use of many genetic markers, it is technically challenging to create isogenic mutants. Therefore, the KPPR1 strain and an isogenic siderophore-negative mutant KP7 were used as positive and negative controls for siderophore-dependent growth, respectively. KPPR1 had the earliest and maximal growth, whereas KP7 was unable to grow (Fig. 3A). The ST258 strains grew but compared to KPPR1, the UHKPC48 strain had a slight delay and NJST258_2, UHKPC05, and VAKPC297 strains had longer delays. Although strains UHKPC48 and VAKPC297 both express the entSΔ 347 allele, the delay in growth of the VAKPC297 isolate is more dramatic than the growth delay of UHKPC48. With the addition of 100 M Fe 2 SO 4 , the growth of KP7 was rescued, and the growth of NJST258_2, UHKPC05, and VAKPC297 improved (Fig. 3B).
Within serum, iron is bound by the host protein transferrin, and iron acquisition for growth by WT KPPR1 requires siderophores (16). To examine the ability of the strains to acquire iron in a physiologically relevant iron-limiting condition, growth curves were performed in M9/Chelex supplemented with 5% heat-inactivated serum. Similar to M9, strain KPPR1 had maximal growth and strain KP7 did not grow. However, all ST258 strains grew rapidly in media containing 5% serum as an iron source, indicating that ST258 strains are able to acquire iron from human serum for replication (Fig. 3C). These data document that all ST258 strains, even those with the entSΔ 347 allele, are able to grow in various iron-depleted and iron-replete growth conditions, although they did not grow as well as KPPR1 strain, which secretes multiple siderophores.
ST258 strains secrete iron-chelating molecules and catechols. To explore the ability of ST258 strains of K. pneumoniae to grow in iron-depleted conditions, in vitro chrome azurol S (CAS) and Arnow assays examining the secretion of iron-chelating molecules were performed. The CAS assay involves a dye bound to iron that can be removed by a chelator, resulting in a shift in the dye color, and is measured as a shift in absorbance at 630 nm (ΔA 630 ) (17). To detect iron-chelating molecules present in bacterial culture after overnight growth in M9 medium, supernatants were subjected to the CAS assay. In addition to KPPR1 and KP7 controls, KP20 was used as a control strain that secretes only Ent based on mass spectrometry (18). Supernatants from all strains, except VAKPC297 (entSΔ 347 ), contained significantly more iron-chelating molecules than KP7, the siderophore negative-control strain, did (Fig. 4A). The lack of secreted iron-chelating molecules by VAKPC297 compared to UHKPC48 is consistent with the growth defect of VAKPC297 in M9/Chelex shown in Fig. 3A. However, the entSΔ 347 strains UHKPC48 and DMC0526 secreted significantly more iron-chelating molecules than strain KP7 did. CAS assay results from strains cultured overnight in M9/Chelex exhibited similar levels of iron-chelating molecules, indicating that residual iron in the M9 minimal medium did not impact the results.
Whereas the CAS assay detects general iron-chelating molecules, the Arnow assay detects catechols, which are the iron coordinating regions of Ent. Similar to the CAS assay, the Arnow assay is a colorimetry-based assay in which a shift in color indicates the presence of catechols (19). To determine whether entSΔ 347 strains secrete catechols, the Arnow assay was performed on bacterium-free culture supernatants. As expected, strains KPPR1 (Ent ϩ Ybt ϩ Sal ϩ ) and KP20 (Ent ϩ ) secreted significantly more catechols than strain KP7 (siderophore negative) did (Fig. 4B). Strains UHKPC05 (WT entS), DMC0526 (entSΔ 347 ), and UHKPC48 (entSΔ 347 ) all secreted significant amounts of catechols, and strain NJST258_2 produced a moderate amount (P ϭ 0.054). Strain NJST258_1 did not secrete significantly more catechols than KP7 despite secreting significantly more iron-chelating molecules (Fig. 4A). Consistent with CAS assay results, strain VAKPC297 (entSΔ 347 ) also did not secrete significantly more catechols than KP7 did. Arnow assay results from strains cultured overnight in M9/Chelex exhibited similar results. These results demonstrate that the majority of ST258 strains are secreting detectable levels of iron-chelating molecules and catechols, despite the entSΔ 347 allele in strains DMC0526 and UHKPC48.
Mass spectrometry detects Ent release by ST258b entS⌬ 347 strains. Because the entSΔ 347 genotype did not correlate consistently with phenotype in the CAS and Arnow assays, we used a more definitive method to measure Ent secretion by ST258 strains of K. pneumoniae. Mass spectrometry has been used to detect siderophores in K. pneu- moniae, independent of their ability to chelate iron or react with Arnow reagents (18). Mass spectrometry analysis revealed that all ST258 strains of K. pneumoniae secreted monomers, dimers, trimers, and cyclic Ent (Fig. 5), even in the absence of a full-length entS. There were no significant differences in the concentrations of monomers, dimers, or overall secretion of Ent products. Compared to the ST258b strain NJST258_2 that contains a full-length copy of entS, entSΔ 347 mutant VAKPC297 produced significantly less cyclic Ent. Despite also harboring this entSΔ 347 mutation, strains DMC0526 and UHKPC48 did not produce significantly less cyclic Ent compared to strain NJST258_2. Strain KPPR1 also secreted less cyclic enterobactin, consistent with conversion of a portion of its Ent to Sal (18). These data confirm that ST258 strains secrete Ent despite the entSΔ 347 allele in some ST258b strains.
CR K. pneumoniae growth in the presence of hemin. RNA-Seq analysis identified that the entSΔ 347 strains UHKPC48 and VAKPC297 produced increased transcripts for hmuS, a hemin transporter, upon growth in M9/Chelexϩserum (Fig. 2B). To determine whether K. pneumoniae strains expressing the entSΔ 347 allele can utilize hemin as an iron source, growth curves were constructed for strains grown in M9/Chelex supplemented with 100, 10, or 1 M hemin (Fig. 6). As in Fig. 3, KPPR1 and ST258 strains were able to grow in M9/Chelex, whereas KP7 (siderophore-negative) strain could not (Fig. 6A). KPPR1 growth improved with hemin addition, whereas KP7 could not grow, indicating that hemin-dependent growth in this strain required siderophores. Addition of 100, 10, and 1 M hemin restored growth in two of three entSΔ 347 strains (DMC0526

FIG 4
The majority of ST258 strains secrete detectable amounts of iron-chelating molecules and catechols. (A and B) To measure the relative concentration of iron-chelating molecules and catechols secreted by each strain, the CAS and Arnow assays were performed, respectively. Data from filtersterilized supernatants from overnight cultures in M9 medium normalized for bacterial growth using OD 600 for each strain from three independent experiments are shown, analyzed using one-way analysis of variance (ANOVA) with Dunnett's posttest to identify significant differences compared to the siderophore-negative strain, KP7. Values significantly different from the value for strain KP7 are indicated by asterisks as follows: **, P Ͻ 0.01; ***, P Ͻ 0.001. Values that are not significantly different (P Ͼ 0.05) from the value for KP7 are indicated (ns). ΔA630, shift in absorbance at 630 nm. and UHKPC48), but not in strain VAKPC297 (Fig. 6B to D), indicating that the hemin may be used as an iron source in some entSΔ 347 strains of ST258 K. pneumoniae during growth in iron-depleted conditions.
Lcn2 inhibits growth of ST258 entS mutants. Mass spectrometry analysis identified that strains are releasing Ent, despite a 347-bp deletion in entS in a subset of ST258b strains (Fig. 5). This Ent release, whether through an active exporter or another indirect method, could promote growth in iron-limited conditions. Growth in serum by K. pneumoniae requires Ent or Sal due to the presence of transferrin as described above (16). To measure growth in serum that is attributable to Ent, the medium was supplemented with Lcn2, an innate immune protein secreted by host epithelial cells and neutrophils that specifically binds to and sequesters Ent (20). Strain KPPR1 that secretes multiple siderophores is not inhibited by the presence of Lcn2 because it secretes Sal,  whereas strain KP20 (Ent ϩ ) is inhibited by Lcn2 (Fig. 7). The siderophore-negative mutant grows poorly regardless of Lcn2, consistent with media supplemented with 10% serum being a siderophore-dependent growth condition. All ST258 strains tested were able to grow in human serum. However, Lcn2 significantly inhibited growth of all ST258 strains, indicating that entSΔ 347 mutants utilize Ent for serum growth (Fig. 7). Although entSΔ 347 mutants expressed the hemin receptor more than other strains, the ability of strains DMC0526 and UHKPC48 to utilize hemin for growth (Fig. 6) did not prevent serum growth inhibition by Lcn2.

DISCUSSION
CR K. pneumoniae is resistant to all or nearly all antibiotics and is a major health care concern (1). One CR MLST type in particular, ST258, has disseminated worldwide. Previous studies have identified epidemiological and genomic differences between two prevalent ST258 clades, clade ST258a and ST258b, including a 347-bp deletion (entSΔ 347 ) in the Ent exporter entS gene (8)(9)(10). This mutation in a highly conserved iron acquisition system suggested differences in iron metabolism among ST258 strains. To examine the transcriptional response of CR K. pneumoniae to iron, we performed RNA-Seq in iron-depleted and iron-replete conditions to identify transcriptional differences. We identified potential mechanisms of iron acquisition by CR K. pneumoniae: the hemin transport operon that has not been characterized previously in K. pneumoniae, a TonB-dependent siderophore receptor FitA, and a catecholate receptor Fiu. We determined that the hemin transport system is expressed at greater levels in entSΔ 347 strains of ST258b K. pneumoniae compared to strains with intact entS. We also characterized the growth patterns and secretion of iron-chelating molecules during in vitro iron-limiting conditions. We found that all ST258 isolates, including entSΔ 347 strains, could grow under iron-limiting conditions and secrete iron-chelating molecules, including Ent. Additionally, we found that hemin stimulates growth of two of three ST258b entSΔ 347 strains tested in iron-depleted media. These findings begin to characterize how CR K. pneumoniae bacteria respond to iron limitation.
Transcriptional analysis by RNA-Seq identified genes that were consistently induced across ST258 strains in response to iron limitation ( Table 2) and those that were differentially regulated in strains with the entSΔ 347 allele. One gene that was upregulated across ST258 isolates upon iron depletion is fitA, a proposed TonB-dependent siderophore receptor (Table 2 and Fig. 1A) (21). In E. coli, the fit operon was found to be an iron acquisition mechanism commonly found in human-pathogenic, but not commensal, clinical isolates (21). The substrate for the fit operon is unknown; however, Ent is not the substrate (21,22). The fit operon represents a mechanism by which ST258 isolates can acquire iron in an Ent-independent manner. A second gene locus of interest that was upregulated in response to iron-depleted conditions by ST258 isolates is the catecholate receptor Fiu (Table 2 and Fig. 1B). The outer membrane receptor Fiu has been hypothesized to mediate uptake of hydrolytic products of Ent, such as linear 2,3-DHBS monomers (23,24). It is possible that ST258 isolates, particularly ST258b isolates that are deficient in a full-length entS, utilize Fiu to scavenge for and take up hydrolytic products of Ent that may have been secreted through mechanisms other than the dedicated Ent exporter. Indeed, mass spectrometry analysis identified that the majority of Ent products secreted by all strains were hydrolytic products of Ent, rather than cyclic Ent (Fig. 5). There were significant differences between strains in the release of cyclic Ent, but no differences in the concentrations of monomers for isolates. This is consistent with prior work illustrating that entS mutants in E. coli secrete 2,3-DHBS (12).
The hemin transport operon was shown to be upregulated by all ST258 isolates in response to iron-depleted conditions. Additionally, it was shown to be significantly upregulated by entSΔ 347 isolates compared to entS ϩ isolates in M9/Chelexϩserum (Fig.  2B). Further in vitro characterization of the ST258 isolates determined that strains UHKPC48 and DMC0526 were able to utilize hemin as an iron source upon growth in iron-depleted media, although strain VAKPC297 was not able to utilize hemin for growth (Fig. 6). However, the ability to utilize hemin for growth did not result in growth of the strains in serum upon the addition of Lcn2 (Fig. 7). One possibility is these strains require Ent for hemin utilization. Indeed, strain KPPR1 requires siderophores to utilize hemin for growth in vitro (Fig. 6), but this strain has low expression of the hemin receptor in M9 medium (Fig. 2). Alternatively, hemin may be scarce in serum, since it can be scavenged by members of the lipocalin superfamily, in particular ␣-1microglobulin (25,26) and was not sufficient to promote growth via hemin transport in the serum growth assay.
The main limitation of this study was that nonisogenic strains were compared to characterize the phenotype associated with the entSΔ 347 allele. We attempted to mitigate this by focusing on comparisons between clonally related ST258b isolates and performing comparative genomics. Despite containing a 347-bp deletion in entS, ST258b isolates UHKPC48 and DMC0526 appear to produce extracellular Ent as measured by the Arnow and CAS assays (Fig. 4). Strain VAKPC297 had detectable Ent, but only when measured by mass spectroscopy. Interestingly, this strain has a nonsynonymous mutation in entC, the gene encoding isochorismate synthase in the Ent synthesis pathway, where an alanine has replaced a glycine at position 362 of the protein (GenBank accession no. EOZ46174). If this affects enzymatic function, it may explain lower detected Ent levels. However, all three strains release sufficient Ent to support robust growth in human serum, a matrix relevant to human infection, and that growth is inhibited by Lcn2 (Fig. 7). It is unclear whether Ent supports growth through a specific secretion mechanism or some other form of release. A more complete understanding of iron acquisition by CR K. pneumoniae requires the examination of additional ST258 clinical isolates and development of efficient tools for isogenic mutant construction and ectopic gene expression in these highly antibiotic-resistant strains. In summary, entSΔ 347 strains have altered hemin receptor expression in response to iron limitation, but they retain the ability to release Ent for growth. Although iron acquisition is essential for virulence, this is a new example of the varied strategies used by K. pneumoniae to carry out a vital function.
K. pneumoniae strains were cultured in LB or M9 minimal medium with and without additional supplementation. M9 minimal salts were diluted and prepared according to the manufacturer's instructions (ThermoFisher, Waltham, MA). To create iron-depleted medium, M9 minimal medium was treated