A study of the correlation between phenotypic antimicrobial susceptibility testing methods and the associated genotypes determined by whole genome sequencing for a collection of Escherichia coli of bovine origin

Abstract

Antimicrobial resistance (AMR) has increased at an alarming pace in the recent years. Molecular-based methods such as whole genome sequencing (WGS) offer a potential alternative to the conventional labour-intensive methods traditionally used to characterise AMR phenotypes. The aim of this study was to investigate whether WGS could be used as a predictor of AMR in Escherichia coli isolates of bovine origin. Genomes of 143 E. coli cultured from cattle presenting with diarrhoea or mastitis were sequenced on an Illumina MiSeq platform. AMR genes were identified using the ResFinder and AMRFinder databases. Antimicrobial susceptibility testing by disk diffusion was performed on a panel of 10 antibiotics, covering 7 antimicrobial classes. Minimum inhibitory concentration (MIC) measurements were made using the Sensititre plate with 6 antibiotics, covering 5 antimicrobial classes. Correlation between genotype and phenotype was assessed statistically by means of a two-by-two table analysis and Cohen's kappa (κ) test. The overall κ correlation between WGS and disk diffusion was 0.81, indicating a near perfect agreement, and the average positive predicted value was 77.4 %. Correlation for individual antimicrobial compounds varied, with five yielding near perfect agreement (κ = 0.81-1.00; amoxicillin, florfenicol, gentamicin, tetracycline and trimethoprim-sulfamethoxazole), one showing substantial agreement (κ = 0.65; nalidixic acid), and four showing moderate agreement (κ = 0.41-0.60). The overall κ correlation between WGS and MIC was 0.55 indicating moderate agreement, and the average positive predicted value was 68.6 %. Three antibiotics yielded near perfect agreement (gentamicin, tetracycline and trimethoprim-sulfamethoxazole) and a further three showed fair agreement (κ = 0.21-0.40). WGS is a useful tool that can be used for the prediction of AMR phenotypes, and correlates well with disk diffusion results. MIC measurements may be necessary for antimicrobial compounds with a high proportion of intermediately resistant isolates recorded, such as cephalothin.

A Retrospective Analysis of Salmonella Isolates across 11 Animal Species (1982–1999) Led to the First Identification of Chromosomally Encoded blaSCO-1 in the USA

“…Slight agreement (k = 0.11) for amoxicillin/clavulanic acid may result from clavulanic acid inhibiting beta-lactamase [ 8 ]. A false positive was observed in nitrofurantoin, possibly due to unknown resistance mechanisms such as the expression of efflux pumps [ 44 , 45 ]. Conversely, a false negative occurred with quinolone-resistant genes ( gyrA(S83F) , parC(T57S) ), indicating phenotypic susceptibility despite the associated genotype, suggesting silenced resistance genes [ 43 ].…”

Antimicrobial resistance and genomic characterization of Salmonella enterica isolates from chicken meat

“…A few numbers of isolates showed false negative results, i.e., isolates that were phenotypically resistant but genotypically susceptible. This is most likely due to the presence of new ARG variants that have not been discovered yet and, therefore, included in the reference database used for prediction ( Neuert et al, 2018 ; Maunsell et al, 2021 ). Another reason for the observed false negative results would be the existence of unknown mechanisms of resistance, such as the expression of efflux pumps and cell wall permeability ( Webber and Piddock, 2003 ; Leon et al, 2018 ; Maunsell et al, 2021 ).…”

“…This is most likely due to the presence of new ARG variants that have not been discovered yet and, therefore, included in the reference database used for prediction ( Neuert et al, 2018 ; Maunsell et al, 2021 ). Another reason for the observed false negative results would be the existence of unknown mechanisms of resistance, such as the expression of efflux pumps and cell wall permeability ( Webber and Piddock, 2003 ; Leon et al, 2018 ; Maunsell et al, 2021 ). No perfect agreement was established between all phenotypically confirmed Salmonella isolates and in silico predicted ARGs.…”