Klebsiella pneumoniae clinical isolates with features of both multidrug-resistance and hypervirulence have unexpectedly low virulence

Abstract

Klebsiella pneumoniae has been classified into two types, classical K. pneumoniae (cKP) and hypervirulent K. pneumoniae (hvKP). cKP isolates are highly diverse and important causes of nosocomial infections; they include globally disseminated antibiotic-resistant clones. hvKP isolates are sensitive to most antibiotics but are highly virulent, causing community-acquired infections in healthy individuals. The virulence phenotype of hvKP is associated with pathogenicity loci responsible for siderophore and hypermucoid capsule production. Recently, convergent strains of K. pneumoniae, which possess features of both cKP and hvKP, have emerged and are cause of much concern. Here, we screen the genomes of 2,608 multidrug-resistant K. pneumoniae isolates from the United States and identify 47 convergent isolates. We perform phenotypic and genomic characterization of 12 representative isolates. These 12 convergent isolates contain a variety of antimicrobial resistance plasmids and virulence plasmids. Most convergent isolates contain aerobactin biosynthesis genes and produce more siderophores than cKP isolates but not more capsule. Unexpectedly, only 1 of the 12 tested convergent isolates has a level of virulence consistent with hvKP isolates in a murine pneumonia model. These findings suggest that additional studies should be performed to clarify whether convergent strains are indeed more virulent than cKP in mouse and human infections.

Fig. 1. Genomic features of antibiotic-resistant clinical isolates of K. pneumoniae from United States hospitals.

A A core genome phylogenetic tree of the 2608 isolates screened in this study. ST, KL type, isolate cohort, and presence of an ESBL or carbapenemase gene are indicated. B Distribution of the number of drug classes per genome for which antimicrobial resistance elements were detected. C The number of genomes that contained each of the different lineages of aerobactin (iuc). In (B) and (C), bars are colored based on the presence of an ESBL or carbapenemase producing enterobacteriaceae (CPE). D The distribution of convergent isolates by sequence type with the different lineages of iuc. cKP- and hvKP-associated STs are labeled in navy blue and red font, respectively. Source data are provided as a Source data file.

Fig. 2. Phylogenetic analysis of convergent, MDR-cKP, NON-MDR-cKp, and hvKP isolates.

A core genome phylogenetic tree of convergent isolates along with representative isolates of MDR-cKP, NON-MDR-cKp, and hvKP groups was generated. The presence of ST, KL type, virulence genes, and ESBL or carbapenemase genes are indicated. Two well-characterized hvKP reference strains are included: NTUH-K2044 and KPPR1. iuc  =  aerobactin biosynthesis genes, iro  =  salmochelin biosynthesis genes, ybt  =  yersiniabactin biosynthesis loci, clb  =  colibactin biosynthesis loci, rmpADC  =  mucoid regulator operon, rmpA2 = mucoid regulator 2 gene, ESBL = extended-spectrum beta-lactamase gene, CPE = carbapenemase producing Enterobacteriaceae.

Fig. 3. Plasmid sequence analysis of convergent, MDR-cKP, hvKP, and NON-MDR-cKP isolates.

A neighbor-joining tree based on mash distance was constructed with all assembled plasmids from MDR-cKP, hvKP, and NON-MDR-cKP (Supplementary Data 3). Plasmid replicons, virulence genes, and classes of AMR genes are indicated. Plasmids from convergent isolates are indicated with green. iuc  =  aerobactin biosynthesis genes, iro  =  salmochelin biosynthesis genes, rmpADC  =  mucoid regulator operon, rmpA2 = mucoid regulator 2 gene, ESBL = extended-spectrum beta-lactamase gene, CPE = carbapenemase producing enterobacteriaceae, AME = aminoglycoside modifying enzyme gene.

Fig. 4. Virulence plasmids identified in 12 representative convergent isolates.

Plasmids from convergent isolates were aligned to A pK2044, a well-characterized hvKP virulence plasmid, or B pEC244_2, an E. coli plasmid. Alignments were made using BRIG with a sequence identity threshold of 85%. Aerobactin biosynthesis genes are indicated with green arrows, salmochelin with orange arrows, rmpA with a blue arrow, and rmpA2 with a red arrow. De novo clustering of C pK2044-related plasmids and D pEC244_2-related plasmids was performed using Mash with a minimum Jaccard index of 0.95. Reference plasmids pK2044 and pEC244_2 were included in the analyses and are indicated. Individual nodes represent an assembled plasmid. Node colors in (C) match plasmid colors in (A), and node colors in (D) match plasmid colors in (B). Tan colored nodes represent plasmids from non-convergent isolates that had Jaccard index of 0.95 or higher with either pK2044 in (C) or pEC244_2 in (D). Networks were graphed using Cytoscape. In panels C and D, antibiotic-resistance genes present on plasmids are listed.

Fig. 5. Phenotypic comparison of convergent, MDR-cKP, hvKP, and NON-MDR-cKP isolates.

A Survival in pooled human serum, B capsule production, C siderophore production, and D hypermucoviscosity as measured by resistance to sedimentation. Each symbol represents an isolate (n = 40). The middle lines represent medians, the boxes represent the interquartile range, and the outer bars represent the minimum and maximum. Data for each isolate are the means of at least 2 independent experiments consisting of 3 technical replicates each. Statistical analysis was performed using one-way ANOVA; B *p = 0.0235, **p = 0.0045, C **p = 0.0076, ^##p = 0.0029, ****p ≤ 0.0001, D ****p ≤ 0.0001. Source data are provided as a Source data file.

Fig. 6. Virulence comparison of convergent, MDR-cKP, hvKP, and NON-MDR-cKP isolates.

Virulence in a murine model of pneumonia was measured for each of the 40 representative isolates. Each symbol represents an isolate (n = 40). The middle lines represent medians, the boxes represent the interquartile range, and the outer bars represent the minimum and maximum. Number of mice and dosing used to determine LD₅₀ values are included in Supplementary Data 5. Statistical analysis was performed using one-way ANOVA; ****p ≤ 0.0001. Source data are provided as a Source data file.