
==== Front
eBioMedicine
EBioMedicine
eBioMedicine
2352-3964
Elsevier

S2352-3964(24)00338-4
10.1016/j.ebiom.2024.105302
105302
Articles
Deciphering the relative importance of genetic elements in hypervirulent Klebsiella pneumoniae to guide countermeasure development
Russo Thomas A. trusso@buffalo.edu
abcd∗
Carlino-MacDonald Ulrike ab
Drayer Zachary J. b
Davies Connor J. ab
Alvarado Cassandra L. ab
Hutson Alan e
Luo Ting L. f
Martin Melissa J. f
McGann Patrick T. f
Lebreton Francois f
a Veterans Administration Western New York Healthcare System, Buffalo, NY, USA
b Department of Medicine, Jacobs School of Medicine and Biomedical Sciences, Buffalo, NY, USA
c Department of Microbiology and Immunology, Jacobs School of Medicine and Biomedical Sciences, Buffalo, NY, USA
d The Witebsky Center for Microbial Pathogenesis, University at Buffalo, State University of New York, Buffalo, NY, USA
e Department of Biostatistics and Bioinformatics, Roswell Park Comprehensive Cancer Center, Buffalo, NY, USA
f Multidrug-Resistant Organism Repository and Surveillance Network (MRSN), Walter Reed Army Institute of Research, Silver Spring, MD, USA
∗ Corresponding author. The Veterans Administration Western New York Healthcare System, 3495 Bailey Ave, Buffalo, NY 14215, USA. trusso@buffalo.edu
22 8 2024
9 2024
22 8 2024
107 10530230 5 2024
7 8 2024
8 8 2024
© 2024 The Author(s)
2024
https://creativecommons.org/licenses/by-nc-nd/4.0/ This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Summary

Background

Quantitating the contribution of phenotype-responsible elements in hypervirulent Klebsiella pneumoniae is needed.

Methods

Isogenic mutants of four hypervirulent clinical isolates that produced K1 (ST23), K2 (ST86), K20 (ST1544), or K54 (ST29) capsules (mean 2.2 log10 LD50 (range 1.5–2.9)) were created to measure the effects on LD50 in a murine model of the hypervirulence-associated plasmid (pVir), iucA, prmpA, prmpA2 (truncated), irp2, and clbBC.

Findings

Curing pVir had the greatest increase in survival (mean LD50 to 7.6 (range 7.0–9.0, p ≤ 0.0001), a dosage comparable to classical K. pneumoniae. Results also showed increased mean LD50s for ΔprmpA (5.9, p ≤ 0.0001), ΔiucA (3.6, p ≤ 0.0001), Δirp2 (3.4), ΔrmpAΔiucA (6.3, p ≤ 0.0001), and ΔpVirΔirp2 (8.7, p ≤ 0.0001). Notably ΔpVir had an additional mean LD50 increase of 1.3 compared to the pVir-encoded ΔprmpAΔiucA (p ≤ 0.01), suggesting presence of additional pVir-virulence genes. Truncated pRmpA2 did not contribute to virulence. Odd ratios in the absence of pVir/yersiniabactin, pVir, pRmpA/aerobactin, pRmpA, aerobactin, yersiniabactin, and colibactin demonstrated a 250-fold, 67-fold, 20-fold, 16.7-fold, 9.6-fold, and 1.7-fold decrease in lethality respectively.

Interpretation

These data can guide countermeasure development.

Funding

This work was supported by NIH R21 AI123558-01 and 1R21AI141826-01A1 (Dr. Russo) and the 10.13039/100000738 Department of Veterans Affairs VA Merit Review (I01 BX004677-01) (Dr. Russo). This study was also partially funded by the U.S. Defense Health Program (DHP) Operations and Maintenance.

Keywords

Klebsiella pneumoniae
Hypervirulent
Virulence factors
Countermeasures
Pathogenesis
Virulence plasmid
==== Body
pmc Research in context

Evidence before this study

Hypervirulent Klebsiella pneumoniae (hvKp) is a bacterium that can cause serious infections in otherwise healthy individuals. Concern has grown with its acquisition of antimicrobial resistance and increasing geographic dissemination. Published data support its hypervirulent pathotype is conferred by a large hvKp-associated virulence plasmid (pVir) and/or genes located on the chromosome. Biosynthetic genes for the iron-acquisition siderophore aerobactin, and prmpA (regulates increased capsule production and mucoviscosity), both located on pVir, have been shown experimentally to contribute to hypervirulence in in vivo infection models. Chromosomal acquisition of integrative and conjugative elements that variably carry genes that produce the siderophore yersiniabactin and the toxin colibactin has been associated with the emergence and dissemination of hvKp. Undoubtedly, factors that contribute to the hypervirulent pathotype remain to be identified.

Added value of this study

Prior to this study there was limited quantitative evidence to determine the relative contribution of these identified factors to hypervirulence. Therefore, a systematic investigation was performed in four representative hvKp clinical isolates via the construction of isogenic derivatives with the loss of pVir, aerobactin, pRmpA, pRmpA2 (truncated), yersiniabactin, and colibactin both alone and in combination, and assessment in a clinically relevant murine infection model. Findings strongly support that pVir is the primary genetic determinant that transforms the baseline virulence potential of classical K. pneumoniae strains to that observed for hvKp strains. Of individual factors, the contribution of pRmpA was > aerobactin, which was > yersiniabactin. Comparison of the loss of pVir to the combined loss of pRmpA and aerobactin supports the existence of additional virulence factors encoded by pVir. Truncated pRmpA2 did not contribute to virulence.

Implications of all the available evidence

Taken together, this study enhances our quantitative understanding of hypervirulent pathotype. These data will enable the development of novel preventative, therapeutic, and control strategies.

Introduction

Hypervirulent Klebsiella pneumoniae (hvKp) is an emerging pathogen of increasing concern. hvKp is significantly more virulent than classical K. pneumoniae (cKp). The pathogenic potential of hvKp is evidenced by its ability to cause tissue invasive, organ and life-threatening infections in healthy hosts from the community. Further, infections with this pathotype often involve multiple sites, some of which are unusual for Enterobacteriaceae, such as the central nervous system, eyes, and fascia.1 Concerningly, some hvKp isolates have become multi- or extensively-drug resistant.2 This combination of hypervirulence and antimicrobial resistance defines a true and dreaded superbug, for which the European Center for Disease Prevention and Control has recently raised an alarm.3,4 To develop strategies for countermeasures directed against hvKp, an increased understanding of the genetic elements and their quantitative importance for increasing virulence is needed. Successful examples in which understanding the biology of infection can lead to countermeasure development is the antimicrobial cefiderocol, a siderophore–antibiotic conjugate that for entry utilises bacterial iron transporters requisite in vivo for iron acquisition and survival,5 and the anti-toxin B monoclonal bezlotoxumab, which reduces recurrent Clostridium difficile infection in high-risk patients.6

hvKp's hypervirulent phenotype is reflected in animal infection models. The LD50 of hvKp after subcutaneous or intraperitoneal injection of outbred mice generally ranges from 101 to 106 CFU, whereas the LD50 of cKp is usually >107 CFU. Current data, which is limited, support that the hypervirulence of hvKp strains is conferred by a large virulence plasmid (pVir; 140- to 220-kilobases)7, 8, 9, 10, 11, 12 or perhaps determinants located on integrative and conjugative elements (ICE).13 Several genes located on pVir have been shown experimentally to contribute to hypervirulence in in vivo infection models, namely: iucA (aerobactin siderophore biosynthesis), prmpA (regulates mucoviscosity phenotype), and peg-344 (putative transporter).14, 15, 16, 17 Chromosomal acquisition of ICEKp10, carrying ybt/irb (yersiniabactin siderophore biosynthesis) and clb (colibactin toxin biosynthesis), has been associated with the emergence and dissemination of the hypervirulent sublineage CG23-I.18 Undoubtedly, unrecognised hvKp specific virulence factors, regulatory factors, and possibly cKp anti-virulence factors remain to be identified. Understanding the relative role of the genes encoded on pVir versus chromosomally located genes will enable a focused search for novel or unrecognised virulence genes and target prioritisation.

In this study, various isogenic mutant constructs were created from four wild-type clinical isolates that produced the most prevalent hvKp capsule types K1 (52–56%), K2 (20–22%), K20 (3–12%), or K54 (5–9%)19, 20, 21 and caused community acquired tissue invasive infection.20 Further, the common and important sequence types ST23 (29%) and ST86 (5%) were represented as were the less common ST29 (2%) and ST1544 (<1%).22 Therefore, these strains represent 4 diverse genetic backgrounds with a balance of common and less common STs, lending confidence that observed effects are likely broadly applicable. These strains were also highly virulent in the validated CD1 SQ challenge murine infection model.20 For all strains, derivatives cured of pVir (ΔpVir) were constructed. Further, when present, isogenic ΔiucA, ΔprmpA, ΔprmpA2, Δirp2, or ΔclbBC derivatives were also created. Lastly for all strains, double ΔpVir/Δirp2 or ΔiucA/ΔprmpA mutants were created. All strains were characterised with the primary goal of identifying high-value targets for countermeasure development. Importantly, optimal countermeasures can be specifically directed against the combination of RmpA, aerobactin, yersiniabactin, and additional pVir-located factors yet to be identified.

Methods

Bacterial strains

The strains used in this study and selected genotypic and phenotypic characteristics are listed in Supplementary Table S2. Four clinical isolates that produced either a K1 capsule (hvKp2, ST23; bacteremia, endophthalmitis; Buffalo, New York, USA), a K2 capsule (hvKp1, ST86; bacteremia, hepatic and splenic abscess; Buffalo, New York, USA), a K20 capsule (hvKp18, ST1544; bacteremia, necrotizing fasciitis left leg; Taipei, Taiwan), or a K54 capsule (hvKp16, ST29; bacteremia, necrotizing fasciitis right neck; Taipei, Taiwan) were the primary strains assessed. For each isolate, constructs were generated in which: 1. the hvKp-associated virulence plasmid (pVir) was cured, 2. iucA (aerobactin synthesis) gene), irp2 (yersiniabactin synthesis), prmpA and prmpA2 (encodes the regulator for mucoid phenotype), and both iucA and prmpA, were deleted, and 3. pVir was cured and irp2 deleted. hvKp2 was the only strain that possessed genes encoding colibactin; therefore, this was the only strain in which clbBC (colibactin synthesis) was deleted, and pVir was cured and clbBC deleted. In three additional randomly chosen strains (hvKp82 (K54, ST29; bacteremia, hepatic abscess; Taipei, Taiwan), hvKp87 (KL2, ST375; bacteremia, hepatic abscess; Oxford, England), and hvKp88 (K20-like, ST893; bacteremia, Montreal, Quebec, Canada)) isogenic mutant derivatives were generated in which only iucA was deleted. Details on strain construction are delineated in the Methods, Supplementary Methods, and Supplementary Tables S3 and S4. All strains were stored at −80 °C prior to use.

Construction of mutant derivatives

The methodology in our laboratory for isogenic mutant construction has evolved over time to enhance efficiency and the likelihood of success. The details for the construction of each mutant derivative are delineated in Supplementary Methods and Supplementary Tables S2 and S3. In brief, for targeting single genes, a DNA fragment that contained a kanamycin resistance cassette flanked by areas of homology with the gene of interest was electroporated into a given strain. Selection for kanamycin resistance was used to identify putative derivatives that underwent allelic exchange. For the creation of mutant constructs involving two genes a kanamycin resistance cassette flanked by flippase recognition target (FRT) sequence was used for the initial gene mutation, followed by flippase (Flp) recombinase-mediated resolution of the kanamycin resistance cassette, and then subsequent mutation of the second gene. The successful deletion of genes was initially confirmed by sequence analysis of a PCR-generated amplicon using primers outside of the gene in question. Curing of the hvKp-associated virulence plasmid was achieved by mutating the plasmid partitioning genes sopAB via allelic exchange, using a kanamycin resistance cassette for selection. The sopAB mutant construct was serially passaged in the absence of kanamycin. Kanamycin sensitive colonies were initially assessed for the loss of the plasmid by the inability to generate PCR-mediated amplicons of plasmid-specific genes. All constructs subsequently underwent whole genome sequencing and analysis to both confirm the gene deletion(s) and/or curing of the virulence plasmid and to exclude cryptic mutations or deletions elsewhere. Both genomic assemblies and raw sequencing data of the isolates used in this study are publicly available in NCBI database under BioProject number PRJNA1090858.

Sequencing and assembly

All strains were sequenced using Illumina MiSeq or NextSeq short read platforms. Short read sequences were assembled de novo using Shovill (v1.0.9) with minimum thresholds for contig size and coverage set at 200 bp and 49.5×, respectively. Additionally, strains hvKp1/2/16/18 were sequenced with PacBio or Minion long read platforms. Long reads were assembled with the Trycycler v0.5.3 pipeline and polished with Illumina short reads to produce a high-quality circularised genome.23

Genotypic typing

Kleborate v2.3.2 was used for in silico K-locus capsule typing.24 A match confidence of ‘high’ and ‘very high’ was used to assign a definitive K-type. A match confidence of ‘good’ was the criteria to append ‘like’ to the best match type. Additionally, Kleborate provided sequence typing for aerobactin, salmochelin, colibactin, yersiniabactin, and the rmpADC loci.

Growth assessment

Growth assays were performed in either lysogeny broth (5 g yeast extract, 10 g NaCl, 10 g tryptone (LB)) or minimal medium (c-M9-CA-te; 200 mL of solution A (2 g (NH4)2SO4, 6 g Na2HPO4, 3 g KH2PO4, 3 g NaCl, 0.011 g Na2SO4), 800 mL of solution B (0.2 g MgCl2, 0.0132 g CaCl2-2H2O, 0.0005 g FeCl3-7H2O, 2.9241 g citrate (trisodium salt dehydrate))), which was subsequently iron-chelated, and then 3 g of casamino acids plus trace elements (5 μg/mL CaCl2, 1 μg/mL CoCl2, 20 μg/mL MgCl2, 10 μg/mL MnCl2) were added. For each experiment, an overnight culture of the strain being assessed was grown at 37 °C in LB broth with isogenic mutant constructs being grown in the presence of kanamycin (40 μg/mL). For LB growth experiments, a fresh culture was inoculated from the overnight culture into LB to an OD600 of approximately 0.09 in 96-well microtiter plates. Plates were incubated with double-orbital shaking at 282 cpm for 24 h at 37 °C in an SynergyH1 or Epoch 2, Biotek, spectrophotometer with OD600 measured in each well every 15 min. For the c-M9-CA-te OD600 growth experiments, 1 mL of an LB overnight culture was washed and re-suspended in 1× PBS before inoculation into the c-M9-CA-te to an OD600 of approximately 0.08 and growth was monitored as described above. A minimum of three biologic repeats with 3 technical repeats was performed for each strain.14

Quantitative mucoviscosity assay

Strains grown overnight in LB were used to inoculate 10 mL of either LB or c-M9-CA-te to a starting OD600 of approximately 0.2. Strains were grown at 37 °C for 24 h, and the assay was performed as described.25 A minimum of 3 biological assays were performed and the results were reported as the mean ± the SD.

Quantitative siderophore assay

Strains were grown overnight at 37 °C in iron-chelated M9 minimal medium containing casamino acids (c-M9-CA)14 and culture supernatants were assessed using the chromeazurol S dye assay as described.17 A minimum of 3 biological assays with 3 technical repeats were performed and the results were reported as the mean ± the SD.

Ethics statement/CD1 mouse subcutaneous (SQ) challenge infection mode

Animal studies were reviewed and approved by the Veterans Administration Institutional Animal Care Committee (IRBNet # 1580121) and the University at Buffalo-SUNY (N/A-Russo1) and were carried out in strict accordance with the recommendations in the guidelines delineated in the “NIH Guide for the Care and Use of Laboratory Animals” (revised 1985) and the “Ethics of Animal Experimentation Statement” (Canadian Council on Animal Care, July 1980) as monitored by the Institutional Animal Care and Use Committee (IACUC). All efforts were made to minimise suffering. Veterinary care for the animals was supplied by the staff of the Veterans Administration Animal Facility under the direction of a fully licensed veterinarian. Studies were performed as described.20 The various challenge inocula and number of mice used for each strain are delineated in Supplementary Table S5. Animals were closely monitored for 14 days after challenge for the study endpoints, which were survival or severe illness (in extremis state/death), which was recorded as a dichotomous variable. Signs that were monitored and which resulted in immediate euthanasia using methods consistent with the recommendations of the American Veterinary Medical Association Guidelines included hunched posture, ruffled fur, labored breathing, reluctance to move, photophobia, and dehydration. The ARRIVE guidelines 2.0 Essential 10 and the Recommended Set of items are addressed in the Supplementary Methods.

Statistical analyses

In vitro growth curves (Supplementary Figures S1 and S2) was analysed as described26; the wild-type parent was compared to its isogenic mutant derivatives via ordinary one-way ANOVA, using Dunnett's multiple comparisons test (DMCT) or paired (PTT) or unpaired (UTT) two-tailed test (Prism 10.2.1 for MacIntosh, GraphPad Software Inc.). Likewise, mucoviscosity when grown in LB or c-M9-CA-te (Fig. 2), as well as siderophore production (Fig. 3a) was assessed via ordinary one-way ANOVA, using DMCT, or PTT or UTT, with the wild-type parent serving as the control for comparison with its mutant derivatives (Prism version 10.2.1).

Pooled analyses for growth, mucoviscosity, and siderophore data were analysed via ordinary one-way ANOVA using DMCT, with the pooled wild-type group (consisting of hvKp1,2,16,18) used as the control for comparison with the pooled mutant groups (Supplementary Table S1, Figs. 2 and 3a). For ΔiucA mutants, two additional pooled cohorts were created (hvKp82,87,88 and hvKp1,2,16,18,82,87,88) for comparing mutants to wild-type using PTTs (Prism 10.2.1; Fig. 4). Isolates hvKp2ΔclbBC and hvKp2ΔpVirΔclbBC were excluded from comparisons to pooled wild-type measurements.

The LD50 was estimated using a logistic regression model as described.2 Pair-wise comparisons of the dose–response curves used to generate LD50 values were used to compare a wild-type parent with a given mutant derivative, pooled comparisons between wild-type strains and mutant derivatives, and odds-ratios and the corresponding 95% confidence intervals and p-values for testing an odds ratio equal to 1 (two-sided) where carried out (SAS/STAT 15.1, Supplementary Table S1, Fig. 3b). Because of the quasi-complete separation in the logistic regression likelihood model when contrasting each mutant with its corresponding wild-type strain due to sparse data across some pairs, we employed a blend of the empirical logit function along with least-squares regression incorporating strain and inoculum factors (CFU/mL) to derive p-values for comparing dose–response curves based on LS-means. Generation of p-values comparing LD50 data was done by a pair-wise comparison of the odds ratios (PWCOR).

Role of funders

The funders had no role in the decision to publish or the preparation of this manuscript. The contents do not represent the views of the U.S. Department of Veterans Affairs or the United States Government. The opinions or assertions contained herein are the private views of the authors and are not to be construed as official or reflecting the views of the Department of the Army or the Department of Defense.

Results

Construction of isogenic mutant derivatives from 4 hvkp clinical isolates

Isogenic mutant derivatives were successfully created from hvKp1(KL2, ST86), hvKp2(KL1, ST23), hvKp16(KL54, ST29), or hvKp18(KL20, ST1544) (summarised in Fig. 1, Supplementary Table S1). All strains were confirmed to be isogenic and to possess the intended genotype based on whole genome sequencing and comparison to their wild-type parent genome.Fig. 1 Relative contributions of genetic elements in mucoviscosity, siderophore production and virulence in a CD1 mice sub-cutaneous model of infection. Heatmaps were generated with the corresponding data from Supplementary Table S1 and organized by strain cohorts a or mutant cohorts b. For each strain or mutant, the presence (green) or absence (grey) of virulence genes carried by pVIR or the chromosome (Chr) are indicated. Growth data in LB or MM are indicated as the area under the curve. Mucoviscosity (Muc.) quantification is indicated as a ratio Post/Pre OD600 from growth in LB or MM. Siderophore production is indicated as μg/mL. Growth, mucoviscosity and siderophore production data are colored with a gradient from 0 (white) to max value (red). The LD50 was estimated using a logistic regression model with the factors for strain and inoculum (CFU/mL). The log LD50 values are colored with a gradient from min (red) to max value (white). Please see methods for details.

In vitro phenotypic studies

Curing pVir or deletion of biomarkers had little to no effect on in vitro growth

All mutant derivatives of isolates used in this study were assessed for growth in LB (nutrient–rich medium) and c-M9-CA-te (nutrient–poor medium) (Fig. 1/Supplementary Figures S1 and S2/Supplementary Table S1a–d,f). No consistent pattern was observed in increased or decreased growth of a given mutant derivative across the 4 strain sets and differences in growth were small on a biological scale. In agreement, pooled data comparing a given mutant derivative in the four strain backgrounds to their wild-type parents demonstrated no significant differences in growth in nutrient-rich or nutrient–poor medium.

Deletion of prmpA and loss of pVir most significantly altered mucoviscosity

The mucoviscosity of hvKp1,2,16,18 and their mutant derivatives were assessed when grown in LB and c-M9-CA-te (Figs. 1 and 2/Supplementary Table S1a-d,f). Quantitatively, 26/34 strains (76%) were more mucoviscous in c-M9-CA-te compared to LB medium. As expected, when grown in either LB or c-M9-CA-te all mutant derivatives deficient in pRmpA (either via deletion of prmpA or curing of pVir) were significantly less mucoviscous than their wild-type parents. All mucoviscosity comparisons to parent strains in rmpA-deficient mutants were statistically significant except for hvKp1ΔpVirΔirp2. Pooled data for the hvKp1,2,16,18 strain sets were consistent with these results.Fig. 2 In vitro quantitative mucoviscosity. In vitro quantitative mucoviscosity data for pooled and individual strain cohorts for hvKp1, hvKp2, hvKp16, and hvKp18 and their isogenic mutant derivatives presented as the mean ± SD. a. Strains grown in LB medium. b. Strains grown in c-M9-CA-te medium. A minimum of three biologic repeats was performed for each strain. Data were analyzed via ordinary one-way ANOVA using Dunnett's multiple comparisons test, with the wild-type parent as the control for comparison with its isogenic mutant derivatives (Prism version 10.2.1). hvKp2ΔclbBC and hvKp2ΔpVirΔclbBC were excluded from pooled analyses. ∗∗∗∗p ≤ 0.0001, ∗∗∗p ≤ 0.001, ∗∗p ≤ 0.01, ∗p ≤ 0.05.

Interestingly, deletion of the truncated prmpA2 in all four wild-type strains resulted in a modest decrease in mucoviscosity when grown in LB medium compared to their wild-type parents; but this decrease was only statistically significant for hvKp16ΔprmpA2 and hvKp18ΔprmpA2. However, by contrast, deletion of prmpA2 resulted in a modest increase in mucoviscosity when grown in c-M9-CA-te medium compared to their wild-type parents, yet this increase was only significant for hvKp16ΔprmpA2. When compared to pooled data, no significant differences were observed in either medium.

Lastly, modest, but significant decreases in mucoviscosity were also observed for hvKp16Δirp2 (40%), hvKp16ΔiucA (33%), and hvKp18ΔiucA (51%) in LB medium. By contrast, significant increases were observed for hvKp1Δirp2 (209%), hvKp1ΔiucA (182%), hvKp16Δirp2 (223%), hvKp16ΔiucA (250%), hvKp18Δirp2 (166%), and hvKp18ΔiucA (160%) in c-M9-CA-te medium. Pooled data demonstrated a significant increase in mucoviscosity for Δirp2 and ΔiucA constructs in c-M9-CA-te medium (both 170%) compared to their wild-type parents.

Curing pVir and deletion of iucA, but not irp2, most significantly altered siderophore production

Quantitative siderophore production of hvKp1,2,16,18 and their mutant derivatives was assessed in c-M9-CA minimal medium (Figs. 1 and 3a/Supplementary Table S1a–d,f). As expected, mutants unable to produce aerobactin (either via deletion of iucA or curing of pVir) produced significantly lower total siderophores than their wild-type parents.16 Pooled data were consistent with these results. Also as expected, a single gene deletion in the yersiniabactin locus of irp2 did not result in a significant decrease in total siderophore when compared to their wild-type parents or with the pooled data analysis of these strains.16Fig. 3 In vitro quantitative siderophore production and in vivo assessment of survival of CD1 mice after sub-cutaneous challenge a. In vitro quantitative total siderophore production data for pooled and individual strain cohorts for hvKp1, hvKp2, hvKp16, and hvKp18 and their isogenic mutant derivatives grown in c-M9-CA medium. Data presented as the mean ± SD. A minimum of three biologic repeats with 3 technical repeats was performed for each strain. Data were analyzed via ordinary one-way ANOVA using Dunnett's multiple comparisons test, with the wild-type parent as the control for comparison with its isogenic mutant derivatives (Prism version 10.2.1). b. Log10 (LD50) values for pooled and individual strain cohorts for hvKp1, hvKp2, hvKp16, and hvKp18 and their isogenic mutant derivatives presented as the mean ± SD. The LD50 was estimated using a logistic regression model with the factors for strain and inoculum (CFU/mL). Pair-wise comparisons of the dose–response curves used to generate LD50 values were used to compare the individual wild-type parents with a given mutant derivative and for pooled comparisons between wild-type strains and mutant derivatives, designated as a pair-wise comparison of the odds ratios [PWCOR] (SAS/STAT 15.1). hvKp2ΔclbBC and hvKp2ΔpVirΔclbBC were excluded from pooled analyses. ∗∗∗∗p ≤ 0.0001, ∗∗∗p ≤ 0.001, ∗∗p ≤ 0.01, ∗p ≤ 0.05.

Interestingly, the deletion of prmpA and prmpA2 variably affected total siderophore production with significant, but modest decreases (hvKp1ΔprmpA (69%), hvKp1ΔprmpA2 (51%), hvKp2ΔprmpA2 (49%), hvKp16ΔprmpA2 (66%)) and increases (hvKp16ΔprmpA (158%), hvKp18ΔprmpA (230%)) compared to their wild-type parents. However, no significant differences were observed with pooled data comparing ΔprmpA or ΔprmpA2 constructs to their wild-type parents.

Curing pVir abolishes the hypervirulent phenotype in a systemic infection model

The virulence of hvKp1,2,16,18 and their mutant derivatives was assessed by challenging outbred CD1 mice subcutaneously to determine at which concentration log10 LD50 (LD50) was achieved. The change in LD50 concentrations attributed to the loss of pVir, pRmpA and pRmpA2, aerobactin, yersiniabactin, colibactin, and selected combinations of these factors were determined (Figs. 1, 3b and 5/Supplementary Table S1a–d,f).

The loss of pVir resulted in the largest decrease in virulence (p ≤ 0.0001 for all strains, [pair-wise comparison of the odds ratios (PWCOR]). The increase in LD50 was 6.1, 5.6, 5.5, and 4.6 for hvKp1ΔpVir, hvKp2ΔpVir, hvKp16ΔpVir, and hvKp18ΔpVir respectively. The pooled data resulted in a mean LD50, of the 4 wild-type strains, of 2.2 ± 0.7 (range 1.5–2.9), whereas curing pVir in these strains resulted in a mean LD50 of 7.6 ± 0.9 (range 7.0–9.0) (p ≤ 0.0001, [PWCOR]); therefore, a mean 5.4 increase in LD50 was attributable to the loss of pVir (Fig. 3b, Supplementary Table S1f). The LD50 of the cured derivatives was similar to the LD50 values observed in classical K. pneumoniae strains (>7),20,25 supporting the concept that the virulence plasmid is the dominant genetic element for conferring the hypervirulent phenotype.

The loss of pRmpA resulted in the next largest decrease in virulence (p ≤ 0.0001 for all strains, [PWCOR]). The increase in LD50 was 4.1, 3.7, 3.5, and 3.5 for hvKp1ΔprmpA, hvKp2ΔprmpA, hvKp16ΔprmpA, and hvKp18ΔprmpA respectively. For pooled data the mean LD50 of the 4 ΔprmpA derivatives was 5.9 ± 0.9 (range 5.2–7.0) (p ≤ 0.0001, [PWCOR]) with a resultant mean 3.7 increase in LD50 being attributable to the loss of pRmpA (Fig. 3b–Supplementary Table S1f). A prmpA2 encoding for a truncated form of pRmpA2 was present in hvKp1,2,16,18; its loss did not decrease virulence.

The impact on LD50 with the loss of aerobactin or yersiniabactin was more variable. The change in LD50 was 3.7, −0.3, 2.5, and −0.4 for hvKp1ΔiucA (p ≤ 0.0001, [PWCOR]), hvKp2ΔiucA (ns), hvKp16ΔiucA (p ≤ 0.001, [PWCOR]), and hvKp18ΔiucA (ns) and −0.5, 1.8, 4.0, and −0.5 for hvKp1Δirp2 (ns), hvKp2Δirp2 (p ≤ 0.01, [PWCOR]), hvKp16Δirp2 (p ≤ 0.0001, [PWCOR]), and hvKp18Δirp2 (ns) respectively. In the pooled data a mean 1.4 and 1.2 increase in LD50 was attributable to the loss of aerobactin (p ≤ 0.0001, [PWCOR]) and yersiniabactin (ns) respectively. In this study, only hvKp2 (KL1, ST23) possessed colibactin. The LD50 of hvKp2ΔclbBC was 1.9 with a resultant 0.4 increase in LD50 being attributable to the loss of colibactin (p ≤ 0.05, [PWCOR]).

Three double mutants were constructed to further assess the impact of specific virulence factors. The loss of yersiniabactin/pVir increased the LD50 by 6.1, 6.4, 7.3, and 6.3 for hvKp1ΔpVirΔirp2, hvKp2ΔpVirΔirp2, hvKp16ΔpVirΔirp2, and hvKp18ΔpVirΔirp2 respectively (p ≤ 0.0001 for all strains, [PWCOR]). For pooled data, the mean LD50 of the 4 ΔpVirΔirp2 derivatives was 8.7 ± 0.5 (range 7.9–9) (p ≤ 0.0001, [PWCOR]) with a resultant mean 6.5 increase in LD50 being attributable to the loss of yersiniabactin/pVir, imparting a near additive effect (mean LD50 attributable to the loss of pVir and yersiniabactin alone was 5.4 and 1.2 respectively).

hvKp2ΔpVirΔclbBC had a LD50 of 7.8 with a resultant 6.3 increase in LD50 attributable to the loss of colibactin/pVir (p ≤ 0.0001, [PWCOR]). Although only a single strain was assessed, this loss suggests a synergistic decrease in virulence (mean LD50 attributable to the loss of pVir and ΔclbBC alone was 5.6 and 0.4 respectively).

Lastly, and insightfully, the loss of both pRmpA and aerobactin (both encoded by genes located on pVir) increased the LD50 6.1, 3.4, 3.6, and 3.5 for hvKp1ΔprmpAΔiucA, hvKp2ΔprmpAΔiucA, hvKp16ΔprmpAΔiucA, and hvKp18ΔprmpAΔiucA respectively (p ≤ 0.0001 for all strains, [PWCOR]). For pooled data the mean LD50 of the four ΔprmpAΔiucA derivatives was 6.3 ± 1.9 (range 4.9–9) (p ≤ 0.0001, [PWCOR]) with a resultant mean 4.2 increase in LD50 being attributable to the loss of pRmpA/aerobactin. For comparison, the resultant mean LD50 attributable to the loss of pRmpA alone and aerobactin alone was 3.7 and 1.4 respectively. This suggests a less than additive effect of decreased virulence that is attributable to the combined loss of pRmpA and aerobactin. When the mean LD50 of pRmpA/aerobactin was compared to the mean LD50 of pVir (mean LD50 of 7.6) there was an additional mean difference of 1.3 in the direction of decreased virulence (p ≤ 0.01, [PWCOR]). These data imply the existence of pRmpA/aerobactin-independent genes located on the hvKp-associated virulence plasmid that also contributes to the hypervirulent phenotype.

Additional in vivo assessment of aerobactin in a systemic infection model

Surprisingly, the individual loss of aerobactin in hvKp2 and hvKp18 did not decrease virulence. Three additional hvKp strains (hvKp82,87,88) were randomly chosen and assessed to better estimate the frequency of this observation (Figs. 1 and 4/Supplementary Table S1e). For selected ΔiucA derivatives, small and likely biologically insignificant decreases in growth were observed, and no significant changes in mucoviscosity were seen when compared to their wild-type parents. Total siderophore production was significantly decreased as expected (p ≤ 0.0001)[unpaired two-tailed t-test]. When evaluated by SQ challenge in the CD1 outbred murine systemic infection model, all isogenic aerobactin-deficient mutants were less virulent than their wild-type parent; the increase in LD50 was 1.9, 4.8, and 2.3 for hvKp82ΔiucA (p ≤ 0.05, [PWCOR]), hvKp87ΔiucA (p ≤ 0.0001, [PWCOR]), and hvKp88ΔiucA (p ≤ 0.05, [PWCOR]) respectively. The mean LD50 for the 3 wild-type strains was 2.6 ± 2.1 (range 1.2–5.0), whereas the loss of aerobactin in these strains resulted in a mean LD50 of 5.6 ± 2.2 (range 3.1–7.2) (p ≤ 0.0001, [PWCOR]); therefore, a mean 3.0 increase in LD50 was attributable to the loss of aerobactin for this strain set. Taken together with data from hvKp1,2,16,18, a mean 2.1 increase in LD50 was attributable to the loss of aerobactin for all 7 strains (p ≤ 0.0001, [PWCOR]). These data demonstrate that aerobactin significantly contributes to the hypervirulent phenotype in 5 of 7 (71%) hvKp strains assessed in this study.Fig. 4 In vitro quantitative mucoviscosity, in vitro quantitative siderophore production, and in vivo assessment of survival of CD1 mice after sub-cutaneous challenge for hvKp82, hvKp87, and hvKp88 with their isogenic aerobactin-deficient mutant derivatives. Data presented for pooled and individual strain cohorts. The (All strains) pooled cohort consists of hvKp1, hvKp2, hvKp16, hvKp18, hvKp82, hvKp87, hvKp88 and their isogenic aerobactin-deficient mutant derivatives. The (82,87,88) pooled cohort consists of hvKp82, hvKp87, hvKp88, and their isogenic aerobactin-deficient mutant derivatives. Data presented as the mean ± SD. a. In vitro quantitative mucoviscosity for strains grown in LB medium. A minimum of three biologic repeats was performed for each strain. b. In vitro quantitative mucoviscosity for strains grown in c-M9-CA-te medium. A minimum of three biologic repeats was performed for each strain. c. In vitro quantitative siderophore production for strains grown in c-M9-CA medium. A minimum of three biologic repeats with 3 technical repeats was performed for each strain. d. Log10 (LD50) values. Mucoviscosity and siderophore data for pooled and individual strain cohorts were analyzed using a paired or unpaired two-tailed t-test, with the ΔiucA mutant derivative group being compared to the corresponding wild-type parent group. The LD50 was estimated using a logistic regression model with the factors for strain and inoculum (CFU/mL). Pair-wise comparisons of the dose–response curves used to generate LD50 values were used to compare the individual wild-type parents with a given mutant derivative and to make pooled comparisons between wild-type strains and mutant derivatives, designated as a pair-wise comparison of the odds ratios [PWCOR] (SAS/STAT 15.1). ∗∗∗∗p ≤ 0.0001, ∗∗∗p ≤ 0.001, ∗∗p ≤ 0.01, ∗p ≤ 0.05.

Relative role of pVir, aerobactin, yesiniabactin, pRmpA, and pRmpA2 in mediating systemic infection after SQ challenge

The relative role of pVir, aerobactin, yesiniabactin, pRmpA, pRmpA2, pVir/yersiniabactin, and pRmpA/aerobactin was further quantitated in this infection model by calculating odds ratios from pooled data (Table 1). For hvKp1,2,16,18 the decrease in lethality in the absence of pVir/yersiniabactin, pVir, pRmpA/aerobactin, pRmpA, aerobactin, and yersiniabactin was 250-fold, 67-fold, 20-fold, 16.7-fold, 9.6-fold, and 1.7-fold respectively (p ≤ 0.0001 for all strains, [PWCOR], except yersiniabactin (ns)). When hvKp82,87,88 were included in the analysis for the loss of aerobactin, there was an 8.9-fold decrease in mortality (p ≤ 0.0001, [PWCOR]). The loss of pRmpA2 increased lethality 8.3-fold (p ≤ 0.05, [PWCOR]).Table 1 Relative role of pVir, aerobactin, yesiniabactin, pRmpA, and pRmpA2 in mediating systemic infection.

Pooled data	95% Wald confidence limits	
Strain	Odds ratio point estimate (WT/Mutant)	Lower confidencelimit	Upper confidencelimit	
ΔpVir
(1,2,16,18)	66.67
∗∗∗∗	31.25	142.9	
ΔprmpA
(1,2,16,18)	16.67
∗∗∗∗	7.463	37.04	
ΔiucA
(1,2,16,18)	9.615
∗∗∗∗	5.236	17.54	
ΔiucA (All Strains)
(1,2,16,18,82,87,88)	8.929
∗∗∗∗	5.405	14.93	
Δirp2
(1,2,16,18)	1.692
ns	0.894	3.205	
ΔprmpA2
(1,2,16,18)	0.127
∗	0.017	0.965	
ΔpVirΔirp2
(1,2,16,18)	250
∗∗∗∗	90.91	500	
ΔprmpAΔiucA
(1,2,16,18)	20
∗∗∗∗	8.475	47.62	
The LD50 was estimated using a logistic regression model with the factors for strain and inoculum (CFU/mL). Pair-wise comparisons of the dose–response curves used to generate LD50 values were used to generate odds-ratios, the corresponding 95% confidence intervals, and p-values for testing an odds ratio equal to 1 (two-sided) (SAS/STAT 15.1). Please see methods for details.

∗∗∗∗p ≤ 0.0001, ∗∗∗p ≤ 0.001, ∗∗p ≤ 0.01, ∗p ≤ 0.05, ns p > 0.05.

Discussion

Several important observations resulted from this study. First, the relative importance of pVir in the four hvKp strains representing different STs and producing different capsule types was determined. When pVir was cured from these strains, the resultant LD50 was on the order of magnitude observed with cKp strains.17,18 These data strongly support that pVir is the primary genetic determinant that transforms the baseline virulence potential of cKp strains to that observed for hvKp strains (Fig. 5).Fig. 5 Relative contributions of genetic elements in hypervirulent Klebsiella pneumoniae in vivo models of infection. A graphic summary (inspired from Choby et al. J Intern Med. 2020 Mar; 287 (3):283–300. https://doi.org/10.1111/joim.13007. PMID: 31677303; PMCID: PMC7057273) of the various virulent determinants and their known characteristics is provided for context. For each wild-type strain or isogenic mutant derivative the log LD50 values observed in sub-cutaneous (SQ) or pulmonary (PUL) CD1 mice model of infection are categorized as log LD50 <3 (red box), log LD50 >3 but <7 (yellow box), or log LD50 >7 (green box). For the SQ model, the pooled log LD50 values from hvKp1,2,16,18 and their isogenic mutant derivatives were used, except data for ΔpVirΔclbBC and ΔclbBC was generated from hvKp2 alone. For the PUL model, the log LD50 values are those of hvKp1 and its isogenic mutant derivatives.

Second, genes located on pVir that encode pRmpA and aerobactin have previously been shown to contribute to the hypervirulent phenotype in systemic infection models.15, 16, 17 To determine the relative contribution of pRmpA and aerobactin, isogenic constructs were created with deletions in prmpA alone, iucA alone, or both prmpA/iucA in each of the four strains. The loss of pRmpA quantitatively increased the LD50 more than the loss of aerobactin. However, the complete loss of pVir imparts an additional 1.3 mean increase in LD50 when compared to the effects of combined pRmpA/aerobactin loss (p ≤ 0.01, [PWCOR]), suggesting there are prmpA/iucA independent genes located on pVir that contribute to the hypervirulent phenotype. This concept is further supported by the difference in odds ratios, a 67-fold versus 20-fold decrease in lethality with the loss of pVir and pRmpA/aerobactin respectively (p = 0.003, [PWCOR]). The identity of the pVir-located genes that could further contribute to the hypervirulent phenotype is unclear. The relative contribution of salmochelin or PEG-344, both encoded on pVir, was not included in this study because prior data in hvKp1 did not demonstrate a role for these factors in systemic virulence mediated by SQ challenge (Fig. 5)14,16; PEG-344 has been shown to contribute to pneumonia via oro-pharyngeal aspiration (Fig. 5).14 However, we cannot rule out a role for these factors in other hvKp strains or infection models.

Third, a prmpA2 with a premature stop codon is commonly observed in hvKp strains; however, it is uncertain whether the truncated pRmpA2 retains functionality. For example, in one study the strain KP309 possessed both a mutated prmpA and prmpA2 but was still highly virulent (LD50 3.3).27 Further, in that same study there were four strains with a wild-type prmpA, but a mutated prmpA2 that remained virulent (LD50 1.8–3.5).27 However, since a corresponding isogenic ΔprmpA2 mutant was not available for comparison, these results do not exclude the possibility that a truncated pRmpA2 could still be functional. In this study, no decrease in virulence was seen in the ΔprmpA2 derivatives of in all four strains (Fig. 5). These findings can be interpreted as the truncated pRmpA2 is not functional, or the truncated pRmpA2 is functional, but its activity is redundant in the presence of a functional pRmpA. The variable, medium dependent changes observed in siderophore production and mucoviscosity with the deletion of prmpA2 suggests the possibility that truncated pRmpA2 could be functional. However, additional studies will be needed to resolve these considerations. It is worth noting that the presence of prmpA2, whether truncated or not, when present together with iucA, iroB, peg-344, and prmpA, has been shown to be predictive of the hvKp pathotype.2,20 If the truncated pRmpA2 is definitively shown to be non-functional, the presence of prmpA2 may confound the effect of unidentified virulence genes in close proximity.

Fourth, although a >1 log10 decrease in virulence was observed with hvKp2Δirp2 (1.8) and hvKp16Δirp2 (4.0) compared to their wild-type parents, hvKp2ΔpVir and hvKp16 ΔpVir had an LD50 consistent with a cKp phenotype (>7). The biosynthetic genes for yersiniabactin are located on the chromosome, usually on integrative conjugative elements (ICE). Further, a bioinformatic screen of an unbiased collection of 3123 MDR K. pneumoniae isolates from global origins revealed that 41.1% of these isolates possessed yersiniabactin biosynthetic genes, usually located on a integrative conjugative element; but only approximately 1.0% of these isolates were predicted to be hvKp based on biomarker count.2 Therefore, taken together, these findings are most consistent with yersiniabactin contributing to the baseline virulence present in cKp strains. Nonetheless, cKp is itself a formidable pathogen and data supports that the presence of yersiniabactin enhances cKp virulence.28 Further, yersiniabactin is more commonly present in strains established to possess the hvKp phenotype.13,18,20,29 These data suggest that in many, but not all hvKp strains, yersiniabactin can enhance virulence. This is supported in this study by an additional increase in LD50 seen with hvKp2ΔpVirΔirp2, hvKp16ΔpVirΔirp2, and hvKp18ΔpVirΔirp2 compared to ΔpVir alone. Importantly, the greatest decrease in lethality, as quantitated by odds ratios, was the combination of the loss of pVir/yersiniabactin (250-fold). These data support that both cKp and hvKp-specific virulence factors need to be targeted for optimal attenuation. The biosynthetic genes for both aerobactin and salmochelin reside on pVir, and although the genes for enterobactin production are chromosomally located, enterobactin is inactivated by lipocalin in vivo.28,30, 31, 32 Therefore, it is tempting to speculate that the inability to produce siderophores active in vivo is the mechanism responsible for decreased virulence seen with the loss of pVir/yersiniabactin. However, data from this report also demonstrated the importance of pRmpA; in fact, its loss decreased mean lethality in all four strains 16.7-fold, more than any other single factor studied. Therefore, it is predicted that countermeasure development targeting yersiniabactin, aerobactin, RmpA and additional unidentified factors encoded on pVir will be needed for maximal attenuation of virulence.

Lastly, previous studies with hvKp1(K2,ST86) demonstrated that aerobactin, but not salmochelin, enterobactin, or yersiniabactin significantly contributed to systemic virulence after SQ challenge.16 These findings raised the question as to whether a specific siderophore (e.g. aerobactin) or the high level of total siderophores mediated by aerobactin or both were requisite for systemic virulence.16 Results from this study provided some insights. The findings observed for hvKp1 were variably true for the other strains; the loss of aerobactin resulted in a significant decrease in virulence for hvKp16 (K54), but not hvKp2(K1) or hvKp18(K20). These findings prompted an assessment of aerobactin in three additional randomly chosen isolates (hvKp82,87,88), and in all these strains the loss of aerobactin resulted in a significant decrease in virulence (1.9, 4.8, 2.3 respectively). Overall, aerobactin alone resulted in a significant decrease in virulence in 71% (5/7) of strains evaluated in this report. Further, in all these strains, the iuc operon was present on the hvKp-associated virulence plasmid (pVir), a chromosomally located iuc operon was not present, and the loss of aerobactin resulted in a significant decrease of total siderophore production, similar to levels seen in cKP strains (except for hvKp16).20 Taken together, these findings support that aerobactin production or a high level of siderophores do not enhance systemic virulence in all hvKp strains after SQ challenge (e.g. hvKp2, hvKp18). It remains unclear which genes could be compensating for the loss of aerobactin in hvKp2 and hvKp18.

A limitation was that only systemic infection mediated by SQ challenge was assessed. It is possible that the relative role of the factors assessed could vary in their contribution to gastrointestinal colonisation, translocation across the intestinal mucosa (if this occurs in humans), pneumonia, or other sites of infection. Another limitation was that only four strains were comprehensively assessed. These strains were chosen because they are well-characterised hvKp isolates that represented the most common capsule types produced by hvKp strains. The common STs 23 and 86 were also represented. However, multi-drug resistant cKp strains that have acquired a partial or complete hvKp-associated virulence plasmid (e.g. ST11-KL64) are worthy candidates for future studies. In addition, the four strains studied possessed the iuc and rmpADC lineage 1; it is possible that different lineages may affect virulence variably.33,34 Similarly, the pVir contained in the 4 strains studied was similar to the canonical hvKp-associated virulence plasmid pLVPK-like/pK2044-like/KpVP-1; however, other hvKp (e.g. ST66, ST380) possess hvKp-associated virulence plasmids with different genomic content, which in turn could variably impact virulence.35

Further, for hvKp1,2,16,18 the loss of yersiniabactin and aerobactin variably affected mucoviscosity in a strain and medium-dependent fashion (similar or decreased in LB, increased in c-M9-CA-te), suggesting a role for iron acquisition with this phenotype. However, the loss of aerobactin did not affect mucoviscosity for hvKp82/hvKp87/hvKp88. Therefore, the significance of this observation, if any, is unclear. Lastly, only hvKp2 possessed colibactin; more data is needed to determine the relative role of colibactin in the pathogenesis of hvKp infection.

In summary, pVir was demonstrated to be the dominant genetic element responsible for the hypervirulent phenotype, that a truncated pRmpA2 does not contribute to systemic virulence, and importantly that countermeasure development directed against yersiniabactin, aerobactin, pRmpA and additional factors encoded on pVir, but not yet identified, will be needed for maximal attenuation of virulence.

Contributors

TR—Conceptualisation, investigation, formal analysis, validation, writing original draft, review & editing, funding acquisition, project administration, resources.

UCM—Investigation, methodology, data curation, review & editing.

ZD—Conceptualisation, investigation, methodology, data curation, formal analysis, review & editing.

CD—Investigation, methodology, data curation, formal analysis, review & editing.

CA—Investigation, methodology, review & editing.

AH—Conceptualisation, methodology, data curation, formal analysis, review & editing.

TL—Investigation, methodology, formal analysis, data curation, review & editing.

MM—Investigation, formal analysis, visualisation, review & editing.

PM—Formal analysis, funding acquisition, review & editing.

FL—Conceptualisation, data curation, formal analysis, visualisation, project administration, review & editing.

All authors reviewed, edited, and approved the manuscript.

Data sharing statement

Both genomic assemblies and raw sequencing data of the isolates analysed in this study are publicly available in NCBI database under the Bio Project number PRJNA1090858.

Declaration of interests

The authors declare that they have no conflicts of interest.

Appendix A Supplementary data

Supplementary Table S1

Detailed in vitro and in vivo phenotypic data for all strains studied.

Supplementary Table S2

Strains and plasmids.

Supplementary Table S3

Primers used in this study.

Supplementary Table S4

hvKp electrocompetent cell growth media, growth conditions and electroporation details.

Supplementary Table S5

CD1 mice subcutaneous challenge data.

Supplementary Methods

Supplementary Figures

Acknowledgements

This work was supported by NIH R21 AI123558-01 and 1R21AI141826-01A1 (Dr. Russo) and the 10.13039/100000738 Department of Veterans Affairs VA Merit Review (I01 BX004677-01) (Dr. Russo). This study was also partially funded by the U.S. Defense Health Program (DHP) Operations and Maintenance. The funders had no role in the decision to publish or the preparation of this manuscript. The contents do not represent the views of the U.S. Department of Veterans Affairs or the United States Government. The authors are thankful to all the staff of the MRSN. The manuscript has been reviewed by the Walter Reed Army Institute of Research and there is no objection to its presentation. The opinions or assertions contained herein are the private views of the authors and are not to be construed as official or reflecting the views of the Department of the Army or the Department of Defense.

Appendix A Supplementary data related to this article can be found at https://doi.org/10.1016/j.ebiom.2024.105302.
==== Refs
References

1 Russo T.A. Marr C.M. Hypervirulent Klebsiella pneumoniae Clin Microbiol Rev 32 3 2019 e00001 e00019 31092506
2 Russo T.A. Alvarado C.L. Davies C.J. Differentiation of hypervirulent and classical Klebsiella pneumoniae with acquired drug resistance mBio 15 2024 e0286723
3 Marr C.M. Russo T.A. Hypervirulent Klebsiella pneumoniae: a new public health threat Expert review of anti-infective therapy 2018
4 European Centre for Disease Prevention and Control Emergence of hypervirulent Klebsiella pneumoniae ST23 carrying carbapenemase genes in EU/EEA countries, first update 2024 ECDC Stockholm Contract No.: TQ-02-24-218-EN-N
5 Echols R. Ariyasu M. Nagata T.D. Pathogen-focused clinical development to address unmet medical need: cefiderocol targeting carbapenem resistance Clin Infect Dis 69 Suppl 7 2019 S559 S564 31724048
6 Johnson S. Gerding D.N. Bezlotoxumab Clin Infect Dis 68 4 2019 699 704 30020417
7 Ye M. Tu J. Jiang J. Clinical and genomic analysis of liver abscess-causing Klebsiella pneumoniae identifies new liver abscess-associated virulence genes Front Cell Infect Microbiol 6 2016 165 27965935
8 Chen Y.T. Chang H.Y. Lai Y.C. Pan C.C. Tsai S.F. Peng H.L. Sequencing and analysis of the large virulence plasmid pLVPK of Klebsiella pneumoniae CG43 Gene 337 2004 189 198 15276215
9 Nassif X. Sansonetti P.J. Correlation of the virulence of Klebsiella pneumoniae K1 and K2 with the presence of a plasmid encoding aerobactin Infect Immun 54 3 1986 603 608 2946641
10 Nassif X. Fournier J.M. Arondel J. Sansonetti P.J. Mucoid phenotype of Klebsiella pneumoniae is a plasmid-encoded virulence factor Infect Immun 57 2 1989 546 552 2643575
11 Tang H.L. Chiang M.K. Liou W.J. Correlation between Klebsiella pneumoniae carrying pLVPK-derived loci and abscess formation Eur J Clin Microbiol Infect Dis 29 6 2010 689 698 20383552
12 Gu D. Dong N. Zheng Z. A fatal outbreak of ST11 carbapenem-resistant hypervirulent Klebsiella pneumoniae in a Chinese hospital: a molecular epidemiological study Lancet Infect Dis 18 1 2018 37 46 10.1016/S1473-3099(17)30489-9 28864030
13 Lin T.L. Lee C.Z. Hsieh P.F. Tsai S.F. Wang J.T. Characterization of integrative and conjugative element ICEKp1-associated genomic heterogeneity in a Klebsiella pneumoniae strain isolated from a primary liver abscess J Bacteriol 190 2 2008 515 526 17981959
14 Bulger J. MacDonald U. Olson R. Beanan J. Russo T.A. Metabolite transporter PEG344 is required for full virulence of hypervirulent Klebsiella pneumoniae strain hvKP1 after pulmonary but not subcutaneous challenge Infect Immun 85 10 2017 e00093 e000917 28717029
15 Cheng H.Y. Chen Y.S. Wu C.Y. Chang H.Y. Lai Y.C. Peng H.L. RmpA regulation of capsular polysaccharide biosynthesis in Klebsiella pneumoniae CG43 J Bacteriol 192 12 2010 3144 3158 20382770
16 Russo T.A. Olson R. MacDonald U. Beanan J. Davidson B.A. Aerobactin, but not yersiniabactin, salmochelin and enterobactin, enables the growth/survival of hypervirulent (hypermucoviscous) Klebsiella pneumoniae ex vivo and in vivo Infect Immun 83 8 2015 3325 3333 26056379
17 Russo T.A. Olson R. Macdonald U. Aerobactin mediates virulence and accounts for increased siderophore production under iron-limiting conditions by hypervirulent (hypermucoviscous) Klebsiella pneumoniae Infect Immun 82 6 2014 2356 2367 24664504
18 Lam M.M.C. Wyres K.L. Duchene S. Population genomics of hypervirulent Klebsiella pneumoniae clonal-group 23 reveals early emergence and rapid global dissemination Nat Commun 9 1 2018 2703 30006589
19 Fang C.T. Lai S.Y. Yi W.C. Hsueh P.R. Liu K.L. Chang S.C. Klebsiella pneumoniae genotype K1: an emerging pathogen that causes septic ocular or central nervous system complications from pyogenic liver abscess Clin Infect Dis 45 3 2007 284 293 17599305
20 Russo T.A. Olson R. Fang C.T. Identification of biomarkers for differentiation of hypervirulent Klebsiella pneumoniae from classical K. pneumoniae J Clin Microbiol 56 9 2018 e00776
21 Yu W.L. Ko W.C. Cheng K.C. Lee C.C. Lai C.C. Chuang Y.C. Comparison of prevalence of virulence factors for Klebsiella pneumoniae liver abscesses between isolates with capsular K1/K2 and non-K1/K2 serotypes Diagn Microbiol Infect Dis 62 1 2008 1 6 18486404
22 Sanikhani R. Moeinirad M. Shahcheraghi F. Molecular epidemiology of hypervirulent Klebsiella pneumoniae: a systematic review and meta-analysis Iran J Microbiol 13 3 2021 257 265 34540163
23 Wick R.R. Judd L.M. Cerdeira L.T. Trycycler: consensus long-read assemblies for bacterial genomes Genome Biol 22 1 2021 266 34521459
24 Lam M.M.C. Wick R.R. Watts S.C. Cerdeira L.T. Wyres K.L. Holt K.E. A genomic surveillance framework and genotyping tool for Klebsiella pneumoniae and its related species complex Nat Commun 12 1 2021 4188 34234121
25 Russo T.A. MacDonald U. Hassan S. An assessment of siderophore production, mucoviscosity, and Mouse infection models for defining the virulence spectrum of hypervirulent Klebsiella pneumoniae mSphere 6 2 2021 e00045 e000421 33762316
26 Russo T.A. Carlino-MacDonald U. Alvarado C.L. Penicillin binding protein 7/8 is a potential drug target in carbapenem-resistant acinetobacter baumannii Antimicrob Agents Chemother 67 1 2023 e0103322
27 Yu W.L. Lee M.F. Tang H.J. Chang M.C. Chuang Y.C. Low prevalence of rmpA and high tendency of rmpA mutation correspond to low virulence of extended spectrum β-lactamase-producing Klebsiella pneumoniae isolates Virulence 6 2 2015 162 172 25830726
28 Bachman M.A. Oyler J.E. Burns S.H. Klebsiella pneumoniae yersiniabactin promotes respiratory tract infection through evasion of lipocalin 2 Infect Immun 79 8 2011 3309 3316 21576334
29 Lam M.M.C. Wick R.R. Wyres K.L. Genetic diversity, mobilisation and spread of the yersiniabactin-encoding mobile element ICEKp in Klebsiella pneumoniae populations Microb Genom 4 9 2018 e000196
30 Bachman M.A. Lenio S. Schmidt L. Oyler J.E. Weiser J.N. Interaction of lipocalin 2, transferrin, and siderophores determines the replicative niche of Klebsiella pneumoniae during pneumonia mBio 3 6 2012 e00224 e00311 23169997
31 Chan Y.R. Liu J.S. Pociask D.A. Lipocalin 2 is required for pulmonary host defense against Klebsiella infection J Immunol 182 8 2009 4947 4956 19342674
32 Flo T.H. Smith K.D. Sato S. Lipocalin 2 mediates an innate immune response to bacterial infection by sequestrating iron Nature 432 7019 2004 917 921 15531878
33 Lam M.M.C. Salisbury S.M. Treat L.P. Genomic and functional analysis of rmp locus variants in Klebsiella pneumoniae bioRxiv 2024 10.1101/2024.05.28.596137
34 Gibbon M.J. Couto N. Cozens K. Convergence and global epidemiology of Klebsiella pneumoniae plasmids harbouring the iuc3 virulence locus bioRxiv 2024 10.1101/2024.01.05.574329
35 Lam M.M.C. Wyres K.L. Judd L.M. Tracking key virulence loci encoding aerobactin and salmochelin siderophore synthesis in Klebsiella pneumoniae Genome Med 10 1 2018 77 30371343
