How Can AI Help Improve Food Safety?

With advances in artificial intelligence (AI) technologies, the development and implementation of digital food systems are becoming increasingly possible. There is tremendous interest in using different AI applications, such as machine learning models, natural language processing, and computer vision to improve food safety. Possible AI applications are broad and include, but are not limited to, (a) food safety risk prediction and monitoring as well as food safety optimization throughout the supply chain, (b) improved public health systems (e.g., by providing early warning of outbreaks and source attribution), and (c) detection, identification, and characterization of foodborne pathogens. However, AI technologies in food safety lag behind in commercial development because of obstacles such as limited data sharing and limited collaborative research and development efforts. Future actions should be directed toward applying data privacy protection methods, improving data standardization, and developing a collaborative ecosystem to drive innovations in AI applications to food safety.

Phylogeographic Clustering Suggests that Distinct Clades of Salmonella enterica Serovar Mississippi Are Endemic in Australia, the United Kingdom, and the United States.

Salmonella enterica serovar Mississippi is the 2nd and 14th leading cause of human clinical salmonellosis in the Australian island state of Tasmania and the United States, respectively. Despite its public health relevance, relatively little is known about this serovar. Comparison of whole-genome sequence (WGS) data of S. Mississippi isolates with WGS data for 317 additional S. enterica serovars placed one clade of S. Mississippi within S. enterica clade B ("clade B Mississippi") and the other within section Typhi in S. enterica clade A ("clade A Mississippi"), suggesting that these clades evolved from different ancestors. Phylogenetic analysis of 364 S. Mississippi isolates from Australia, the United Kingdom, and the United States suggested that the isolates cluster geographically, with U.S. and Australian isolates representing different subclades (Ai and Aii, respectively) within clade A Mississippi and clade B isolates representing the predominant S. Mississippi isolates in the United Kingdom. Intraclade comparisons suggested that different mobile elements, some of which encode virulence factors, are responsible for the observed differences in gene content among isolates within these clades. Specifically, genetic differences among clade A isolates reflect differences in prophage contents, while differences among clade B isolates are due to the acquisition of a 47.1-kb integrative conjugative element (ICE). Phylogenies inferred from antigenic components (fliC, fljB, and O-antigen-processing genes) support that clade A and B Mississippi isolates acquired these loci from different ancestral serovars. Overall, these data support that different S. Mississippi phylogenetic clades are endemic in Australia, the United Kingdom, and the United States. IMPORTANCE The number of known so-called "polyphyletic" serovars (i.e., phylogenetically distinct clades with the same O and H antigenic formulas) continues to increase as additional Salmonella isolates are sequenced. While serotyping remains a valuable tool for reporting and monitoring Salmonella, more discriminatory analyses for classifying polyphyletic serovars may improve surveillance efforts for these serovars, as we found that for S. Mississippi, distinct genotypes predominate at different geographic locations. Our results suggest that the acquisition of genes encoding O and H antigens from different ancestors led to the emergence of two Mississippi clades. Furthermore, our results suggest that different mobile elements contribute to the microevolution and diversification of isolates within these two clades, which has implications for the acquisition of novel adaptations, such as virulence factors.

Transcriptional profiling of the L. monocytogenes PrfA regulon identifies six novel putative PrfA-regulated genes.

The transcriptional activator Positive Regulatory Factor A (PrfA) regulates expression of genes essential for virulence in Listeria monocytogenes. To define the PrfA regulon, the 10403S wildtype (WT) strain, a constitutively active prfA* mutant, and an isogenic ∆prfA mutant were grown under PrfA-inducing conditions in a medium containing glucose-1-phosphate and pre-treated with 0.2% activated charcoal. RNA-seq-generated transcript levels were compared as follows: (i) prfA* and WT; (ii) WT and ∆prfA and (iii) prfA* and ∆prfA. Significantly higher transcript levels in the induced WT or constitutively active PrfA* were identified for 18 genes and 2 ncRNAs in at least one of the three comparisons. These genes included: (i) 10/12 of the genes previously identified as directly PrfA-regulated; (ii) 2 genes previously identified as PrfA-regulated, albeit likely indirectly; and (iii) 6 genes newly identified as PrfA-regulated, including one (LMRG_0 2046) with a σA-dependent promoter and PrfA box located within an upstream open reading frame. LMRG_0 2046, which encodes a putative cyanate permease, is reported to be downregulated by a σB-dependent anti-sense RNA. This newly identified overlap between the σB and PrfA regulons highlights the complexity of regulatory networks important for fine-tuning bacterial gene expression in response to the rapidly changing environmental conditions associated with infection.

Contributions of σ(B) and PrfA to Listeria monocytogenes salt stress under food relevant conditions.

Listeria monocytogenes is well known to survive and grow under several stress conditions, including salt stress, which is important for growth in certain foods as well as for host infection. To characterize the contributions, to salt stress response, of transcriptional regulators important for stress response and virulence (i.e., σ(B) and PrfA), we analyzed three L. monocytogenes parent strains and isogenic mutants (ΔsigB, ΔprfA, and ΔsigBΔprfA), representing different serotypes and lineages, for their ability to grow, at 25°C, in BHI with 1.9 M NaCl. With regard to growth rate, only the lineage IV strain presented a significant difference between the parent strain and both of its respective mutants lacking prfA (ΔprfA and ΔsigBΔprfA). Conversely, the lineage I and II parent strains showed significantly shorter lag phase in comparison to their respective ΔsigB mutant strains. Intestinal epithelial cell invasion assay and hemolytic activity assays showed a significant role for σ(B) in the former and for PrfA in the latter. To explore the mechanism that may contribute to the extended lag phase in the ΔsigB mutant strain and survival and growth of the parent strain upon salt shock, whole genome transcription profiling was performed to compare transcript levels between the lineage I, serotype 1/2b, parent strain and its isogenic ΔsigB mutant after 30 min of lag phase growth at 25°C in the presence of 1.9M NaCl (salt shock) without aeration. Microarray data showed significantly higher transcript levels for 173 genes in the parent strain as compared to the ΔsigB strain. Overall, 102 of the 173 σ(B) up-regulated genes had been identified in previous studies, indicating that 71 genes were newly identified as being up-regulated by σ(B) in this study. We hypothesize that, among these genes newly identified as σ(B) up-regulated, four genes (lmo2174, lmo0530, lmo0527 and lmo0529) may play a major role in response to salt stress. Lmo2174 contains domains that facilitate sensing and producing a transduction signal in the form of cyclic di-GMP, which may activate the enzymes Lmo0527, Lmo0529 and Lmo0530, which encode proteins similar to those responsible for synthesis of exopolysaccharides that may protect the cell by changing the cell wall structure during salt stress. Overall, our data showed that σ(B), but not PrfA, contributes to growth under salt stress. Moreover, we show that the σ(B) regulon of a L. monocytogenes lineage I strain challenged with salt shock includes salt stress-specific as well as previously unidentified σ(B) up-regulated genes.

σ(B) plays a limited role in the ability of Listeria monocytogenes strain F2365 to survive oxidative and acid stress and in its virulence characteristics.

Listeria monocytogenes strain F2365 was the first strain representative of serotype 4b (lineage I) to be sequenced in 2004, suggesting it could become the model organism for this serotype, which is associated with most human outbreaks of listeriosis worldwide to date. F2365 itself is an outbreak strain that was involved in the listeriosis outbreak associated with Mexican-style soft cheese in California in 1985. In this study, we show through phenotypic and transcriptomic analysis that L. monocytogenes strain F2365 has reduced ability to respond to acid and oxidative stress. F2365 has neither the σ(B)-dependent ability to survive acid or oxidative stress nor the σ(B)-dependent ability to infect Caco-2 epithelial cells in vitro or guinea pigs in vivo. More studies are needed to determine whether the atypical σ(B)-independent response to stress observed in F2365 is strain specific, serotype specific, or even lineage specific.

Refinement of the Listeria monocytogenes σB regulon through quantitative proteomic analysis.

σ(B) is an alternative σ factor that regulates stress response and virulence genes in the foodborne pathogen Listeria monocytogenes. To gain further insight into σ(B)-dependent regulatory mechanisms in L. monocytogenes, we (i) performed quantitative proteomic comparisons between the L. monocytogenes parent strain 10403S and an isogenic ΔsigB mutant and (ii) conducted a meta-analysis of published microarray studies on the 10403S σ(B) regulon. A total of 134 genes were found to be significantly positively regulated by σ(B) at the transcriptomic level with >75 % of these genes preceded by putative σ(B)-dependent promoters; 21 of these 134 genes were also found to be positively regulated by σ(B) through proteomics. In addition, 15 proteins were only found to be positively regulated by σ(B) through proteomics analyses, including Lmo1349, a putative glycine cleavage system protein. The lmo1349 gene is preceded by a 5' UTR that functions as a glycine riboswitch, which suggests regulation of glycine metabolism by σ(B) in L. monocytogenes. Herein, we propose a model where σ(B) upregulates pathways that facilitate biosynthesis and uptake of glycine, which may then activate this riboswitch. Our data also (i) identified a number of σ(B)-dependent proteins that appear to be encoded by genes that are co-regulated by multiple transcriptional regulators, in particular PrfA, and (ii) found σ(B)-dependent genes and proteins to be overrepresented in the 'energy metabolism' role category, highlighting contributions of the σ(B) regulon to L. monocytogenes energy metabolism as well as a role of PrfA and σ(B) interaction in regulating aspects of energy metabolism in L. monocytogenes.

Listeria monocytogenes {sigma}B has a small core regulon and a conserved role in virulence but makes differential contributions to stress tolerance across a diverse collection of strains.

Listeria monocytogenes strains are classified in at least three distinct phylogenetic lineages. There are correlations between lineage classification and source of bacterial isolation; e.g., human clinical and food isolates usually are classified in either lineage I or II. However, human clinical isolates are overrepresented in lineage I, while food isolates are overrepresented in lineage II. sigma(B), a transcriptional regulator previously demonstrated to contribute to environmental stress responses and virulence in L. monocytogenes lineage II strains, was hypothesized to provide differential abilities for L. monocytogenes survival in various niches (e.g., food and human clinical niches). To determine if the contributions of sigma(B) to stress response and virulence differ across diverse L. monocytogenes strains, DeltasigB mutations were created in strains belonging to lineages I, II, IIIA, and IIIB. Paired parent and DeltasigB mutant strains were tested for survival under acid and oxidative stress conditions, Caco-2 cell invasion efficiency, and virulence using the guinea pig listeriosis infection model. Parent and DeltasigB mutant strain transcriptomes were compared using whole-genome expression microarrays. sigma(B) contributed to virulence in each strain. However, while sigma(B) contributed significantly to survival under acid and oxidative stress conditions and Caco-2 cell invasion in lineage I, II, and IIIB strains, the contributions of sigma(B) were not significant for these phenotypes in the lineage IIIA strain. A core set of 63 genes was positively regulated by sigma(B) in all four strains; different total numbers of genes were positively regulated by sigma(B) in the strains. Our results suggest that sigma(B) universally contributes to L. monocytogenes virulence but specific sigma(B)-regulated stress response phenotypes vary among strains.

Recombination and positive selection contribute to evolution of Listeria monocytogenes inlA.

The surface molecule InlA interacts with E-cadherin to promote invasion of Listeria monocytogenes into selected host cells. DNA sequencing of inlA for 40 L. monocytogenes isolates revealed 107 synonymous and 45 nonsynonymous substitutions. A frameshift mutation in a homopolymeric tract encoding part of the InlA signal peptide was identified in three lineage II isolates, which also showed reduced ability to invade human intestinal epithelial cells. Phylogenies showed clear separation of inlA sequences into lineages I and II. Thirteen inlA recombination events, predominantly involving lineage II strains as recipients (12 events), were detected and a number of amino acid residues were shown to be under positive selection. Four of the 45 non-synonymous changes were found to be under positive selection with posterior probabilities >95 %. Mapping of polymorphic and positively selected amino acid sites on the partial crystal structure for InlA showed that the internalin surface of the leucine-rich repeat (LRR) region that faces the InlA receptor E-cadherin does not include any polymorphic sites; all polymorphic and positively selected amino acids mapped to the outer face of the LRR region or to other InlA regions. The data show that (i) inlA is highly polymorphic and evolution of inlA involved a considerable number of recombination events in lineage II isolates; (ii) positive selection at specific amino acid sites appears to contribute to evolution of inlA, including fixation of recombinant events; and (iii) single-nucleotide deletions in a lineage II-specific 3' homopolymeric tract in inlA lead to complete loss of InlA or to production of truncated InlA, which conveys reduced invasiveness. In conclusion, inlA has a complex evolutionary history, which is consistent with L. monocytogenes' natural history as an environmental pathogen with broad host-range, including its adaptation to environments and hosts where different inlA alleles may provide a selective advantage or where inlA may not be required.