WGS-Based Prediction and Analysis of Antimicrobial Resistance in Isolates From Israel.

Rapid developments in the field of whole genome sequencing (WGS) make antimicrobial resistance (AMR) a target within reach. is a leading cause of foodborne infections in Israel with increasing rates of resistance. We applied WGS analysis to study the prevalence and genetic basis of AMR in 263 human and veterinary representative isolates retrieved from a national collection during 2003-2012. We evaluated the prediction of phenotypic AMR from genomic data. Genomes were screened by the NCBI AMRFinderPlus and the BioNumerics tools for acquired AMR genes and point mutations. The results were compared to phenotypic resistance determined by broth microdilution. The most prevalent resistant determinants were the multi-drug efflux transporter gene (100%), the tetracycline resistance gene (82.1%), the quinolone resistance g T861 point mutation (75.7%), and the streptomycin resistance gene. A variety of 12 known β lactam resistance genes ( variants) were detected in 241 (92%) isolates, the most prevalent being , , and (56, 16, and 7%, respectively). Other aminoglycoside resistance genes and the macrolide resistance point mutation were rare (<1%). The overall correlation rate between WGS-based genotypic prediction and phenotypic resistance was 98.8%, sensitivity, specificity, positive, and negative predictive values being 98.0, 99.3, 99.1, and 98.5%, respectively. wgMLST-based phylogeny indicated a high level of clonality and clustering among the studied isolates. Closely related isolates that were part of a genetic cluster (single linkage distance ≤ 15 alleles) based on wgMLST phylogeny mostly shared a homogenous AMR determinant profile. This was observed in 18 of 20 (90.0%) clusters within clonal complex-21, suggesting clonal expansion of resistant isolates. Strong association to lineage was noted for the gene and the various genes. High resistance rates to tetracycline and quinolones and a low resistance rate to macrolides were detected among the Israeli isolates. While a high genotypic-phenotypic correlation was found, some resistance phenotypes could not be predicted by the presence of AMR determinants, and particularly not the level of resistance. WGS-based prediction of antimicrobial resistance in requires further optimization in order to integrate this approach in the routine workflow of public health laboratories for foodborne surveillance.