Logo: to the web site of Uppsala University

Antibiotic heteroresistance (HR) is a phenotype characterized by the presence of low frequency subpopulations with increased resistance present in a more susceptible main population of bacteria. The clinical impact of the phenotype is not clear, but in vitro, in vivo and clinical studies suggest that HR increases the risk of treatment failure. To further complicate the situation, HR often evades detection by commonly used diagnostic methods. The phenotype can be caused by several different genetic mechanisms, and one main mechanism is tandem gene amplification of resistance genes. In this thesis, questions regarding clinical implications, prevalence, diagnostics and genetic mechanisms are adressed in three papers.

In Paper I, the prevalence, classification and clinical outcomes of breakpoint crossing HR (BCHR) in 255 Escherichia coli bloodstream infection isolates were investigated for three clinically relevant antibiotics in a retrospective study. The BCHR prevalences for cefotaxime, gentamicin and piperacillin-tazobactam were <1%, 43% and 9%, respectively. Out of 125 BCHR isolates 96% (120/125) were classified as suceptible by disk diffusion. Patients with BCHR infections that were treated with the corresponding antibiotic had higher odds for admittance to the intensive care unit and mortality for gentamicin, and for admittance to the intermediate care unit for piperacillin-tazobactam.

In Paper II, we predicted the HR phenotype from whole genome sequencing data utilizing machine learning algorithms. 467 clinical isolates of E. coli phenotyped for piperacillin-tazobactam HR were included. The best performing model detected HR isolates with 100% sensitivity and 84.6% specificity. Genetic analysis of the resistant subpopulations showed that copy number increases of resistance genes, either due to amplifications or increased plasmid copy number, were the main HR mechanisms.

In Paper III, the population distribution of resistance gene tandem amplifications in a HR E. coli isolate was resolved and studied, using a new method combining genetic engineering and Nanopore long read sequencing. The distribution of amplifications correlated with the observed HR phenotype. Mathematical modelling suggested that indirect resistance mechanisms could affect the distribution of amplification copy numbers.  

In conclusion, these findings advance the understanding of the prevalence, clinical outcome, diagnostics and genetic mechanisms of the HR phenotype. The presented methodologies in Paper II and III can aid in developing better diagnostics for detection of HR, and in further investigations of the parameters that shapes the HR population and how these populations impact the clinical outcomes of antibiotic treatment.