A Practical Method to Implement Strain-Level Metagenomics-Based Foodborne Outbreak Investigation and Source Tracking in Routine

Abstract

The management of a foodborne outbreak depends on the rapid and accurate identification of the responsible food source. Conventional methods based on isolation of the pathogen from the food matrix and target-specific real-time polymerase chain reactions (qPCRs) are used in routine. In recent years, the use of whole genome sequencing (WGS) of bacterial isolates has proven its value to collect relevant information for strain characterization as well as tracing the origin of the contamination by linking the food isolate with the patient's isolate with high resolution. However, the isolation of a bacterial pathogen from food matrices is often time-consuming and not always successful. Therefore, we aimed to improve outbreak investigation by developing a method that can be implemented in reference laboratories to characterize the pathogen in the food vehicle without its prior isolation and link it back to human cases. We tested and validated a shotgun metagenomics approach by spiking food pathogens in specific food matrices using the Shiga toxin-producing Escherichia coli (STEC) as a case study. Different DNA extraction kits and enrichment procedures were investigated to obtain the most practical workflow. We demonstrated the feasibility of shotgun metagenomics to obtain the same information as in ISO/TS 13136:2012 and WGS of the isolate in parallel by inferring the genome of the contaminant and characterizing it in a shorter timeframe. This was achieved in food samples containing different E. coli strains, including a combination of different STEC strains. For the first time, we also managed to link individual strains from a food product to isolates from human cases, demonstrating the power of shotgun metagenomics for rapid outbreak investigation and source tracking.

Keywords: SNP analysis; STEC; food surveillance; metagenomics; outbreak; whole genome.

Figure 1

Presentation of 5 different workflows for the preparation of metagenomics samples of spiked beef (light blue) and the conventional method for Shiga toxin-producing Escherichia coli (STEC) detection and characterization based on several steps of qPCR and isolation on selective media (ISO/TS 13136:2012, grey). The extracted DNA (amplified or not, from metagenomics samples or isolate) is tested for quality control (DNA purity, integrity, concentration) before sequencing on the Illumina MiSeq in parallel to a qPCR check for the presence of stx genes (green).

Figure 2

Presentation of the bioinformatics analysis for the characterization of STEC in samples using a metagenomics approach. After sequencing and pre-processing of the reads, first, the species in the sample are detected by a taxonomic classification tool (Kraken2); then, the presence of a pathogen in the sample is predicted based on the detection of virulence genes in the reads (SRST2), after which individual bacterial strains are inferred (Sigma) and characterized with gene detection (SRST2) and single nucleotide polymorphism (SNP) phylogeny (SNP-calling pipeline).

Figure 3

Percentages of reads classified to the genus level using Kraken2 (taxonomic classification tool) from beef samples with in-house databases of mammals, archaea, bacteria, fungi, human, protozoa, and viruses. Light blue represents the proportion of “Bos” corresponding to beef reads. Yellow represents the presence of “Escherichia” in the sample. The reads that could not be classified to the genus level for mammals, archaea, bacteria, fungi, human, protozoa, or viruses are represented in gray. (A) Blank meat samples; Bk-0h—non-enriched blank; BK-24h—non-spiked meat sample enriched for 24 h, 1–3 biological replicates. (B) Extraction kits; workflow A—Nucleospin Food, workflow B—DNeasy Blood & Tissue and workflow C—Zymo HostZERO. (C) Enrichment times; workflow A—24 h culture enrichment, workflow D—16 h culture enrichment, workflow E—16 h culture enrichment, extraction followed by DNA amplification using phi 29 DNA polymerase; all extracted with Nucleospin Food kit. (D) Biological and technical replicates of workflow A. Small differences in the detected species shown in panels A, C, and D can be explained by the heterogeneity of the samples and biological variation, as different replicates of the experiment were used.

Figure 4

Gene depth per million trimmed reads per sample for the detection of genes encoding for serotype O157:H7 (wzx and fliC genes) and the stx1a, stx2a, eae, and ehxA virulence genes (5 genes from ISO/TS 13136:2012 and ehxA present on plasmid pO157) with more than 80% query coverage and 80% identity in all reads for beef samples processed with different workflows A-B-C-D-E, and in biological (A-1, A-2, A-3) and technical replicates (A1-3, A2-3, A3-3) of workflow A. Increasing depth (per million trimmed reads) is represented in shades of green to yellow according to the color gradient in the legend.

Figure 5

(A) SNP-based phylogenetic tree of STEC strains inferred from metagenomics samples (dark blue) and of sequenced isolates (black). Reference: E. coli O157:H7 str. Sakai (BA000007.2). Beef/goat isolate: STEC isolate obtained after following the conventional method on the prepared spiked samples. (B) Phylogenetic tree of the STEC O157 with percentage of the reference genome covered and gene detection in the strains. Orange: closely related strains from the outbreak cluster. (C) Phylogenetic tree of the STEC O103 with percentage of the reference genome covered and gene detection in the strains. Blue: closely related strains. (D) Phylogenetic tree of the STEC O145 with percentage of the reference genome covered and gene detection in the strains. Green: closely related strains. The scale bar represents nucleotide substitution per 100 nucleotide site. Node values represent bootstrap support values.

Figure 6

(A) Percentages of reads classified to the genus level using Kraken2 (taxonomic classification tool) on all reads of goat cheese samples with in-house databases of mammals, archaea, bacteria, fungi, human, protozoa, and viruses. Green represents the proportion of “Capra” corresponding to goat reads. Yellow represents the presence of “Escherichia” in the sample. The reads that could not be classified to the genus level for mammals, archaea, bacteria, fungi, human, protozoa, or viruses are represented in gray. (B) Gene depth per million trimmed reads per sample of wzx and fliC genes for the determination of types O103, O145, H2, and H28 and stx1a, eae, and ehxA virulence genes with more than 80% coverage and 80% identity in all reads of goat cheese samples. Increasing depth (per million trimmed reads) is represented in shades of green to yellow according to the color gradient in the legend. Goat_Bk_24h = Blank goat cheese enriched for 24 h. Goat_O103 = goat cheese spiked with STEC O103. Goat_O145 = goat cheese spiked with STEC O145. Goat_O103+O145 = goat cheese co- spiked with STEC O103 and STEC O145.