The emergence of highly resistant and hypervirulent Klebsiella pneumoniae CC14 clone in a tertiary hospital over 8 years

Abstract

Background: Klebsiella pneumoniae is a major bacterial and opportunistic human pathogen, increasingly recognized as a healthcare burden globally. The convergence of resistance and virulence in K. pneumoniae strains has led to the formation of hypervirulent and multidrug-resistant strains with dual risk, limiting treatment options. K. pneumoniae clones are known to emerge locally and spread globally. Therefore, an understanding of the dynamics and evolution of the emerging strains in hospitals is warranted to prevent future outbreaks.

Methods: In this study, we conducted an in-depth genomic analysis on a large-scale collection of 328 multidrug-resistant (MDR) K. pneumoniae strains recovered from 239 patients from a single major hospital in the western coastal city of Jeddah in Saudi Arabia from 2014 through 2022. We employed a broad range of phylogenetic and phylodynamic methods to understand the evolution of the predominant clones on epidemiological time scales, virulence and resistance determinants, and their dynamics. We also integrated the genomic data with detailed electronic health record (EHR) data for the patients to understand the clinical implications of the resistance and virulence of different strains.

Results: We discovered a diverse population underlying the infections, with most strains belonging to Clonal Complex 14 (CC14) exhibiting dominance. Specifically, we observed the emergence and continuous expansion of strains belonging to the dominant ST2096 in the CC14 clade across hospital wards in recent years. These strains acquired resistance mutations against colistin and extended spectrum β-lactamase (ESBL) and carbapenemase genes, namely bla_OXA-48 and bla_OXA-232, located on three distinct plasmids, on epidemiological time scales. Strains of ST2096 exhibited a high virulence level with the presence of the siderophore aerobactin (iuc) locus situated on the same mosaic plasmid as the ESBL gene. Integration of ST2096 with EHR data confirmed the significant link between colonization by ST2096 and the diagnosis of sepsis and elevated in-hospital mortality (p-value < 0.05).

Conclusions: Overall, these results demonstrate the clinical significance of ST2096 clones and illustrate the rapid evolution of an emerging hypervirulent and MDR K. pneumoniae in a clinical setting.

Keywords: Klebsiella pneumoniae; Antimicrobial resistance; Precision epidemiology; Whole-genome sequencing.

Fig. 1

Phylogenetic tree of the collection and the metadata. The neighbor-joining tree for 328 genomes under study. The tree was reconstructed from the pairwise sequence (SNP) distance values, by ape package in R. The SNPs were called with Snippy (see “Methods”). We rooted the tree using midpoint rooting. The resistance profile shows the distribution of major resistance determinants, i.e., Carb (carbapenemase), ESBL, AGly (aminoglycosides), Flq (fluroquinolone), and Tet (tetracyclines) (see Supplemental Table S1 for the full list of resistance determinants). The virulence and resistance scores were computed from Kleborate, based on the presence of key virulence and resistance determinants (see “Methods”) [27]. Numbers on the distributions of virulence score and count of resistance genes and classes and resistance score show the absolute value. The terms “mut” and “acq” stand mutations and acquired resistance genes, respectively. The coverage strips show the percentage of sites of the plasmid sequences to which short reads for each strain were mapped

Fig. 2

Genotypic convergence of resistance and virulence. The panel corresponds to genomes with and without ESBL or carbapenemase genes. The bubbles correspond to the distribution of genome harboring virulence genes for the siderophores of aerobactin (iuc) and yersiniabactin (ybt). The genes were identified using Kleborate (see “Methods”). The shaded region shows the strains with convergence of virulence and resistance as per the score computed by Kleborate, based on the virulence and resistance gene profiles

Fig. 3

The phylodynamic of ST2096 strains. A The Bayesian phylodynamic tree for the ST2096 strains. The horizontal red lines show 95% Highest Posterior Distribution (HPD) for the age of the internal node. We showed the distribution of carbapenemase, ESBL positive, with the acquired genes for aminoglycoside and the distribution of missense, stop gained, and frameshift mutations in the mgr gene and the acquired mcr gene. Colors on the patient panel show whether the isolates were retrieved from the same patients. We did not show the results for patients with one representative strain. B Bayesian skyline plot for the size of the population over time. The shaded region corresponds to 95% confidence interval. C Ancestral state reconstruction for the presence of the carbapenemase genes for the ST2096 clone. The pie charts represent the likelihood of the ancestral state. The tree is a Bayesian tree in (A). The tree tips show the presence/absence status of the genes. The clade colors show the BAPS clusters in (A). D Contextualization of the ST2096 clone with the global population. The neighbor-joining tree was reconstructed from the SNP distance matrix for the ST2096 cluster. Five divergent strains were removed before reconstructing the tree

Fig. 4

Transmission dynamics of ST2096 clone. A Inferred transmission network of ST2096 in the hospital under study (for indirect transmission). Nodes and edges represent patients and indirect transmission routes with a probability more significant than the cutoff of 0.1, respectively. The thickness of the edges corresponds to the probability transmission. Patients who survived until the end of hospital treatment are marked with blue circles. Other patients had a deceased status. B The distribution of patient status, body site, and admission hospital ward in networks with different cutoffs for the transmission routes. C The relationship between the total node degree and betweenness centrality with hospitalization length, measured in days. The red line shows the fitted linear regression, and the shade shows 95% confidence interval. For a definition of degree and betweenness centrality, refer to “Methods”

Fig. 5

Integration of ST data with clinical and demographic data. A The odds ratio of death for the major STs. The error bars show a 95% confidence interval. The dotted line shows an odds ratio of one, above which the presence of STs is positively linked with death. The asterisks ** and *** correspond to p-values of < 0.01 and < 0.001 from Z-test, respectively. The black and orange colors for the asterisks show the significance when the odds ratio smaller and bigger in ST2096 than the other group, respectively. B The coefficient value associated with STs from logistic regression analysis in which sex, gender, age, and site of isolation were included as predictors. The coefficient is the expected change in log odds of being an ST type. C Distribution of patients’ age and body site of infection. The ** shows p-value < 0.01 from Wilcoxon signed-rank test

Fig. 6

Integration of clinical data for patients with ST data. A Charlson comorbidity index and length of hospitalization (LOH) across patients. The index is computed from comorbidities, as detailed in the methods. Length of hospitalization (LOH) in days across in-patients. B Relative frequency of patient conditions linked with infections across patient groups carrying different STs, indicated in ICD-10 codes. The codes A41, R05, R07, and R50 correspond to sepsis, cough, pain in throat and chest, and fever, respectively. The horizontal lines in A, B, and C represent the significance between ST2096 and other STs. The statistical test for A and B was the Wilcoxon signed-rank test, and for C, it was the one-sided proportion test. C The mean term importance measure (tf-idf) for the topmost important words overrepresented in the diagnostic tests for ST2096. Error bars show 95% confidence interval from 100 bootstrapped samples. The * and ** signs correspond to a significance level of < 0.05 and < 0.01, respectively, based on the Wilcoxon signed-rank test. The black and orange colors for the asterisks correspond to cases where the mean value (in A and B) and relative frequency (in C) for ST2096 are greater and smaller than the other group, respectively