Differentiation of hypervirulent and classical Klebsiella pneumoniae with acquired drug resistance

Differentiation of hypervirulent and classical Klebsiella pneumoniae with acquired drug resistance

17 January 2024 | Thomas A. Russo, Cassandra L. Alvarado, Connor J. Davies, Zachary J. Drayer, Ulrike Carlino-MacDonald, Alan Hutson, Ting L. Luo, Melissa J. Martin, Brendan W. Corey, Kara A. Moser, Kamile Rasheed, Alison L. Halpin, Patrick T. McGann, Francois Lebreton
This study aimed to differentiate hypervirulent (*Klebsiella pneumoniae*) (hvKp) from classical (*K. pneumoniae*) (cKp) strains, particularly those with acquired drug resistance. The authors assembled 49 *K. pneumoniae* strains with combinations of *iucA*, *iroB*, *peg-344*, *rmpA*, and *rmpA2* genes and evaluated their virulence using a murine infection model. They assessed various biomarkers, including biomarker count, siderophore production, mucoviscosity, Kleborate virulence score, and Mash/Jaccard distances to the canonical pLVKP plasmid. Both logistic regression and a CART model were used to determine the most predictive variables. The results showed that the biomarker count was the strongest predictor, with an area under the curve of 0.962 for logistic regression and 94% accuracy for the CART model. The presence of all five biomarkers was most accurate, while a count of ≥4 was 100% sensitive but less specific and accurate. These findings provide valuable information for identifying hvKp strains, which is crucial for clinical care, surveillance, and research.This study aimed to differentiate hypervirulent (*Klebsiella pneumoniae*) (hvKp) from classical (*K. pneumoniae*) (cKp) strains, particularly those with acquired drug resistance. The authors assembled 49 *K. pneumoniae* strains with combinations of *iucA*, *iroB*, *peg-344*, *rmpA*, and *rmpA2* genes and evaluated their virulence using a murine infection model. They assessed various biomarkers, including biomarker count, siderophore production, mucoviscosity, Kleborate virulence score, and Mash/Jaccard distances to the canonical pLVKP plasmid. Both logistic regression and a CART model were used to determine the most predictive variables. The results showed that the biomarker count was the strongest predictor, with an area under the curve of 0.962 for logistic regression and 94% accuracy for the CART model. The presence of all five biomarkers was most accurate, while a count of ≥4 was 100% sensitive but less specific and accurate. These findings provide valuable information for identifying hvKp strains, which is crucial for clinical care, surveillance, and research.
Reach us at info@study.space