A diverse array of micro-organisms can be found on food, including those that are pathogenic or resistant to antimicrobial drugs. Metagenomics involves extracting and sequencing the DNA of all micro-organisms on a sample, and here, we used a combination of culture and culture-independent approaches to investigate the microbial ecology of food to assess the potential application of metagenomics for the microbial surveillance of food. We cultured common foodborne pathogens and other organisms including Escherichia coli, Klebsiella/Raoultella spp., Salmonella spp. and Vibrio spp. from five different food commodities and compared their genomes to the microbial communities obtained by metagenomic sequencing following host (food) DNA depletion. The microbial populations of retail food were found to be predominated by psychrotrophic bacteria, driven by the cool temperatures in which the food products are stored. Pathogens accounted for a small percentage of the food metagenome compared to the psychrotrophic bacteria, and cultured pathogens were inconsistently identified in the metagenome data. The microbial composition of food varied amongst different commodities, and metagenomics was able to classify the taxonomic origin of 59% of antimicrobial resistance genes (ARGs) found on food to the genus level, but it was unclear what percentage of ARGs were associated with mobile genetic elements and thus transferable to other bacteria. Metagenomics may be used to survey the ARG burden, composition and carriage on foods to which consumers are exposed. However, food metagenomics, even after depleting host DNA, inconsistently identifies pathogens without enrichment or further bait capture.

Yersinia enterocolitica is an underreported cause of foodborne gastroenteritis. Little is known of the diversity of Y. enterocolitica isolated from food and which food commodities contribute to human disease. In this study, Y. enterocolitica was isolated from 37/50 raw chicken, 8/10 pork, 8/10 salmon and 1/10 leafy green samples collected at retail in the UK. Up to 10 presumptive Y. enterocolitica isolates per positive sample underwent whole genome sequencing (WGS) and were compared with publicly available genomes. In total, 207 Y. enterocolitica isolates were analyzed and belonged to 38 sequence types (STs). Up to five STs of Y. enterocolitica were isolated from individual food samples and isolates belonging to the same sample and ST differed by 0-74 single nucleotide polymorphisms (SNPs). Biotype was predicted for 205 (99 %) genomes that all belonged to biotype 1A, previously described as non-pathogenic. However, around half (51 %) of food samples contained isolates belonging to the same ST as previously isolated from UK human cases. The closest human-derived isolates shared between 17 and 7978 single nucleotide polymorphisms (SNPs) with the food isolates. Extensive food surveillance is required to determine what food sources are responsible for Y. enterocolitica infections and to re-examine the role of biotype 1A as a human pathogen.

Campylobacter spp. are leading bacterial gastroenteritis pathogens. Infections are largely underreported, and the burden of outbreaks may be underestimated. Current strategies of testing as few as one isolate per sample can affect attribution of cases to epidemiologically important sources with high Campylobacter diversity, such as chicken meat. Multiple culture method combinations were utilized to recover and sequence Campylobacter from 45 retail chicken samples purchased across Norwich, UK, selecting up to 48 isolates per sample. Simulations based on resampling were used to assess the impact of Campylobacter sequence type (ST) diversity on outbreak detection. Campylobacter was recovered from 39 samples (87%), although only one sample was positive through all broth, temperature, and plate combinations. Three species were identified (Campylobacter jejuni, Campylobacter coli, and Campylobacter lari), and 33% of samples contained two species. Positive samples contained 1-8 STs. Simulation revealed that up to 87 isolates per sample would be required to detect 95% of the observed ST diversity, and 26 isolates would be required for the average probability of detecting a random theoretical outbreak ST to reach 95%. An optimized culture approach and selecting multiple isolates per sample are essential for more complete Campylobacter recovery to support outbreak investigation and source attribution.

All foods carry microbes, many of which are harmless, but foods can also carry pathogens and/or microbial indicators of contamination. Limited information exists on the co-occurrence of microbes of food safety concern and the factors associated with their presence. Here, a population-based repeated cross-sectional design was used to determine the prevalence and co-occurrence of Escherichia coli, Klebsiella spp., Salmonella spp. and Vibrio spp. in key food commodities -chicken, pork, prawns, salmon and leafy greens. Prevalence in 1,369 food samples for these four target bacterial genera/species varied, while 25.6% of all samples had at least two of the target bacteria and eight different combinations of bacteria were observed as co-occurrence profiles in raw prawns. Imported frozen chicken was 6.4 times more likely to contain Salmonella than domestic chicken, and imported salmon was 5.5 times more likely to be contaminated with E. coli. Seasonality was significantly associated with E. coli and Klebsiella spp. contamination in leafy greens, with higher detection in summer and autumn. Moreover, the odds of Klebsiella spp. contamination were higher in summer in chicken and pork samples. These results provide insight on the bacterial species present on foods at retail, and identify factors associated with the presence of individual bacteria, which are highly relevant for food safety risk assessments and the design of surveillance programmes.

Non-typhoidal Salmonella (NTS) is a major cause of bacterial gastroenteritis. Although many countries have implemented whole genome sequencing (WGS) of NTS, there is limited knowledge on NTS diversity on food and its contribution to human disease. In this study, the aim was to characterise the NTS genomes from retail foods in a particular region of the UK and assess the contribution to human NTS infections. Raw food samples were collected at retail in a repeated cross-sectional design in Norfolk, UK, including chicken (n=311), leafy green (n=311), pork (n=311), prawn (n=279) and salmon (n=157) samples. Up to eight presumptive NTS isolates per positive sample underwent WGS and were compared to publicly available NTS genomes from UK human cases. NTS was isolated from chicken (9.6 %), prawn (2.9 %) and pork (1.3 %) samples and included 14 serovars, of which Salmonella Infantis and Salmonella Enteritidis were the most common. The S. Enteritidis isolates were only isolated from imported chicken. No antimicrobial resistance determinants were found in prawn isolates, whilst 5.1 % of chicken and 0.64 % of pork samples contained multi-drug resistant NTS. The maximum number of pairwise core non-recombinant single nucleotide polymorphisms (SNPs) amongst isolates from the same sample was used to measure diversity and most samples had a median of two SNPs (range: 0-251). NTS isolates that were within five SNPs to clinical UK isolates belonged to specific serovars: S. Enteritidis and S. Infantis (chicken), and S. I 4,[5],12:i- (pork and chicken). Most NTS isolates that were closely related to human-derived isolates were obtained from imported chicken, but further epidemiological data are required to assess definitively the probable source of the human cases. Continued WGS surveillance of Salmonella on retail food involving multiple isolates from each sample is necessary to capture the diversity of Salmonella and determine the relative importance of different sources of human disease.

Food products carry bacteria unless specifically sterilised. These bacteria can be pathogenic, commensal or associated with food spoilage, and may also be resistant to antimicrobials. Current methods for detecting bacteria on food rely on culturing for specific bacteria, a time-consuming process, or 16S rRNA metabarcoding that can identify different taxa but not their genetic content. Directly sequencing metagenomes of food is inefficient as its own DNA vastly outnumbers the bacterial DNA present. We optimised host DNA depletion enabling efficient sequencing of food microbiota, thereby increasing the proportion of non-host DNA sequenced 13-fold (mean; range: 1.3-40-fold) compared to untreated samples. The method performed best on chicken, pork and leafy green samples which had high mean prokaryotic read proportions post-depletion (0.64, 0.74 and 0.74, respectively), with lower mean prokaryotic read proportions in salmon (0.50) and prawn samples (0.19). We show that bac-terial compositions and concentrations of antimicrobial resistance (AMR) genes differed by food type, and that salmon metagenomes were influenced by the production/harvesting method. The approach described in this study is an efficient and effective method of identifying and quantifying the predominant bacteria and AMR genes on food.

Consumption of prawns as a protein source has been on the rise worldwide with seafood identified as the predominant attributable source of human vibriosis. However, surveillance of non-cholera Vibrio is limited both in public health and in food. Using a population-and market share-weighted study design, 211 prawn samples were collected and cultured for Vibrio spp. Contamination was detected in 46 % of samples, and multiple diverse Vibrio isolates were obtained from 34 % of positive samples. Whole genome sequencing (WGS) and phylogenetic analysis illustrated a comprehensive view of Vibrio species diversity in prawns available at retail, with no known pathogenicity markers identified in Vibrio parahaemolyticus and V. cholerae. Antimicrobial resistance genes were found in 77 % of isolates, and 12 % carried genes conferring resistance to three or more drug classes. Resistance genes were found predominantly in V. parahaemolyticus, though multiple resistance genes were also identified in V. cholerae and V. vulnificus. This study highlights the large diversity in Vibrio derived from prawns at retail, even within a single sample. Although there was little evidence in this study that prawns are a major source of vibriosis in the UK, surveillance of non-cholera Vibrio is very limited. This study illustrates the value of expanding WGS surveillance efforts of non-cholera Vibrios in the food chain to identify critical control points for food safety through the production system and to determine the full extent of the public health impact.