Introduction: The development of rapid diagnostics is key to combating multidrug-resistant (MDR) Enterobacteriaceae. Whole genome sequencing (WGS) offers rapid genetic characterisation of MDR bacteria, however, prediction of antimicrobial resistance (AMR) phenotype of Enterobacteriaceae from WGS data remains a challenge. Here we investigate the genetic diversity of MDR clinical Escherichia coli isolates from Australian livestock and assess the phenotypic predictive capabilities of WGS.
Methods: 44 E. coli isolates with multidrug resistance and/or resistance to extended-spectrum-cephalosporins (ESC) were selected from 324 isolates collected between 2013-2014 based on a weighted score from Vitek® minimum inhibitory concentration assays. Further phenotypic detection tests were performed on all suspect ESC resistant isolates. In silico analysis of Illumina reads and assembled genome data were used to determine the sequence type (ST), AMR genotype and virulence gene profile of each isolate.
Results: Reliable prediction of AMR phenotype from genotype using WGS for tetracyclines, aminoglycosides and folate-pathway inhibitors was shown. Sensitivity values for predictive power was greater than 0.96 for all drugs tested with the exceptions of amoxicillin-clauvanic acid and ESCs. ESC resistance could be predicted in 7 isolates by extended spectrum or ampC beta-lactamase carriage: blaCTX-M-14 (n=3), blaCTX-M-9 (n=1), or blaCMY-2-like (n=3), respectively. Of 22 ESC-resistant isolates (cefpodoxime ≥8 mg/L) that did not carry an extended spectrum or ampC beta-lactamase gene, 15 harboured mutations within the chromosomal ampC promotor region that has been shown to cause AmpC hyperproduction. We also identified a new ampC promoter mutation and a suspected insertion sequence-mediated ampC induction. In four cases, we could not predict the ESC phenotype from genotype.