Abordagem Integrada para Predição de Perfis de Resistência a Antimicrobianos à Partir de Dados Genômicos de Isolados Clínicos de Corynebacterium spp.

Abstract

Antimicrobial resistance represents a threat to global public health. The rapid spread of multidrug-resistant pathogenic bacteria and the inability of antibiotics to effectively combat them have led to more severe and prolonged infections. To mitigate this challenge, whole-genome sequencing has been proposed as a tool for epidemiological surveillance and as an alternative to phenotypic tests for predicting resistance, with the promise of rapid, accurate results and better-targeted treatment. However, its application for the latter faces significant obstacles. For clinically important organisms such as Mycobacterium tuberculosis, some studies can establish up to 99% concordance between genotype and antibiotic susceptibility phenotype. On the other hand, other microorganisms, such as emerging pathogens of the Corynebacterium genus, are far from achieving an acceptable level of concordance in clinical application, reflecting a lack of understanding of the microorganism and its resistance mechanisms. In this work, we sequenced the multidrug-resistant isolate VH4248 from the Hospital Marqués de Valdecillla in Santander, Spain, identified as C. urealyticum by classical microbiological and VITEK1 methods. Genome-based identification tools established only a 93.7% similarity with the C. urealyticum database, making reclassification to Corynebacterium sp. necessary. We also assembled, annotated, and explored the genomes of 107 publicly available C. striatum isolates using the PATRIC (BV-BRC), KmerResistance, and Resfinder tools for genomic prediction of antibiotic resistance to beta-lactams, aminoglycosides, macrolides, tetracyclines, and fluoroquinolones. We compared the results with phenotypic data through ROC curves and identified the widely distributed ermX, tetW, and blaA genes with ROC curve areas (AUC) > 0.874 for erythromycin, tetracycline, and penicillin antibiotics. However, all prediction tools showed suboptimal capacity for predicting resistance in C. striatum when evaluated according to international parameters: major error rates (MERs) and very major error rates (VMERs). Accordingly, when we evaluated new clinical isolates of Corynebacterium spp. (n = 18) by polymerase chain reaction (PCR) using primer oligonucleotides for the major identified antibiotic resistance genes, only tetW and ermX genes showed good levels of concordance with phenotypic antibiotic susceptibility test results (94.1% and 83.3%, respectively). Therefore, we conclude that further studies are needed to standardize and establish efficient strategies for resistance prediction based on genomic data from clinically relevant Corynebacterium spp.