FOOD SAFETY IN THE ERA OF NEXT-GENERATION SEQUENCING: GENOMIC CHARACTERIZATION OF SHIGA TOXIN-PRODUCING ESCHERICHIA COLI AND METAGENOMIC SURVEILLANCE OF IRRIGATION SURFACE WATER

Abstract

In this study, we first utilized high-throughput next-generation sequencing (NGS) and bioinformatic analyses to characterize potential public health threats posed by non-top-7 Shiga toxin-producing Escherichia coli (STEC). NGS allowed us to detect virulence (n = 46) and antimicrobial resistance (AMR) (n = 27) factors within the genomes of the STEC strains, to make genome-wide comparisons with published human clinical isolates, and to characterize three novel O-antigen gene clusters. We found that the distribution of 33 virulence genes and 15 AMR determinants exhibited significant differences among serotypes (P < 0.05), and that 47 strains were genetically related to human clinical strains inferred from a pan-genome phylogenetic tree. We secondly developed a web tool, PhyloPlus, that allowed users to generate customized bacterial and archaeal phylogenies, which can be incorporated into their own microbial community studies. We also utilized two public datasets (human microbiome, n = 60; fermented food metagenomes, n = 62) to illustrate how application of phylogeny improved our analyses. We showed that the integration of phylogenies introduced alternative phylogeny-based diversity metrics and allowed more conservative null model constructions, thereby reducing potential inflation of type I errors. Finally, we employed deep metagenomic shotgun sequencing, and our developed web tool, to investigate on a collection of 404 surface water samples collected from four regions in Latin America. We reported the high detection rates of pathogenic and contaminant bacteria in these samples, including Salmonella (29.21%), Listeria (6.19%), and E. coli (35.64%), necessitating the monitoring and proper treatment on these surface waters. We also described the regional differences in terms of sample taxonomic composition and the resistome, and further presented key factors that drove the separation patterns for each sampling region. We utilized recent metagenomic assembly and binning algorithms to report the construction of 1,461 de-replicated metagenome-assembled genomes (MAGs) that were of at least medium quality. The incorporation of the MAGs into the taxonomic classifier Kraken2’s database led to a 12.85% increase in classifiable sequence reads. Additionally, we conducted network analysis on AMR genes and the genus-level taxonomy, based on assembled contigs, to provide information to better understand the dynamics of the transferring of AMR genes.