Genomic insights into antimicrobial resistance and isolation source classification of Cronobacter sakazakii from infant and toddler food in the United States.

Cronobacter sakazakii is an opportunistic foodborne pathogen associated with severe infections in infants, linked to powdered infant formula (PIF) and related products. We conducted genomic profiling of C. sakazakii (n=209) from infant and toddler food in the United States, comprising all publicly available genomes for this source, through the integration of antimicrobial resistance (AMR) gene (ARG), plasmid replicon, virulence gene, phylogenetic, and pan-genome analyses. We further applied a machine learning (ML)-driven isolation source classification approach based on pan-genome features to distinguish food and clinical isolates. AMR analysis revealed a conserved resistome dominated by three β-lactam resistance genes (blaCSA, blaCSA-1, and blaCSA-2). Independent co-occurrence and pairwise association analysis of ARGs and plasmid replicons indicated sparse and gene-specific relationships, suggesting that observed AMR patterns were more consistent with conserved resistance determinants than extensive plasmid-mediated dissemination. Phylogenetic analysis identified two major clades, while pan-genome assessment demonstrated an open genome dominated by accessory genes. Using gene presence/absence profiles, a random forest classifier achieved high accuracy in distinguishing food and clinical isolates, highlighting the classification power of pan-genome signatures within the dataset. These findings provide insights into the genomic structure of food-associated C. sakazakii and the utility of integrating comparative genomics with ML for food safety surveillance.