Abstract
Bloodstream infections (BSIs) and urinary tract infections (UTIs), predominantly caused by Escherichia coli and Klebsiella spp., are an increasing clinical challenge. Rapid empiric antimicrobial therapy is vital to prevent UTIs from progressing to BSIs and BSIs progressing to sepsis. This study aims to improve empiric antimicrobial prescribing, which in the UK is currently based on historic phenotypic data collected at regional level.This study investigates the diversity and molecular epidemiology of deduplicated, sequential E. coli from BSIs (n=669) and UTIs (n=199), as well as Klebsiella spp. from BSIs (n=210), isolated at a regional diagnostic laboratory serving a population of 1.5 million people including Bristol, Bath, North Somerset and South Gloucestershire. All isolates were subjected to whole genome sequencing (WGS) and data were interrogated for antimicrobial resistance (AMR) determinants and phylogenetic relationships. Genotypic data were compared with phenotypic antimicrobial susceptibility data provided by the diagnostic laboratory, and the concordance with predictions from commonly used methods for AMR prediction based on WGS was evaluated.
There was a significant rise in E. coli BSIs in the summer of 2020, however this was not due to any particular source of infection or E. coli sequence type (ST). It was found that E. coli ST131 BSIs were significantly biased towards a urinary tract source of infection, and that carriage of TEM genes by BSI E. coli was more common if the infection came from a urinary source. Phylogenetic analysis showed resistance mechanisms were dispersed and intermixed among isolates from different phylogroups. Comparison between BSI 2018 and 2020 show significant increases in TEM, OXA and ampC genes. Significant increases were observed with genes associated with resistance to fluoroquinolones in UTIs between 2018 and 2020. Analysis of highly resistant ST15 K. pneumoniae containing blaNDM-1 indicated transmission in the same hospital ward. K. pneumoniae BSI isolates from two different patients showed 100% identity for AMR and virulence regions on plasmids having IncFIB, IncR and Col440 replicons. Concordance for genomic AMR prediction varied across bioinformatics tools for both E. coli and Klebsiella spp. Lowest critical errors were detected across all bioinformatics tools for cefotaxime. No critical errors were observed with Kleborate for ciprofloxacin.
In silico detection and characterisation of AMR via WGS to inform empiric therapy is still in its infancy and requires further optimisation before utilisation alone in clinical settings. However, characterisation of infection source, AMR and virulence genes in BSIs and UTIs will be important factors in optimising empiric therapy.
| Date of Award | 23 Jan 2024 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Matthew B Avison (Supervisor) |