Recent proposed poultry regulation by USDA Food Safety and Inspection Service (FSIS) shows a change in the approach to controlling NTS, with an increased focus on higher risk serovars based on their epidemiology, underlying genetic virulence, and concerns for their greater public health risks. Anticipating future food safety threats and future regulatory targets across species requires deeper understanding of the population dynamics of NTS serovars and the genes they carry. Evaluating how Salmonella populations evolve and emerge requires a complex combination of genetic, epidemiological, and statistical methods. The objectives of this study were 1) to use genomics, machine learning, and epidemiological approaches to identify Salmonella serovars in meats that are of greatest public health concern, and genes most associated with virulence in humans and 2) to assess temporal dynamics in NTS serovar prevalence and virulence factor gene profiles.
To accomplish objective 1, this study improved and expanded upon methods initially developed by this group to create a scalable approach that can be updated over time and can also be applied to other pathogens. Using the National Center for Biotechnology Information (NCBI) Pathogen Detection Isolate Browser (The NCBI Pathogen Detection Project, 2016) researchers collected 64,284 NTS isolate genomes sourced from beef, chicken, and human clinical cases, along with pertinent metadata, including isolation date. To more accurately determine each NTS isolate’s VF profile, this research expanded upon previous assemblage of VFs, adding VFs from the UniProt database and expanding the database to include VFs from 32 different pathogenic genera (The UniProt Consortium, 2024). To investigate connections between VF types and virulence, VFs were grouped into eight categories, e.g., exotoxin or nutrition, expanding upon the categorization methods used by the virulence factor database (VFDB). Isolates were annotated using a protein clustering approach to determine their VF carriage. Two hierarchical Bayesian models were used to determine the relative risk (multiplier) per serving for different serovars, relative to their presence in meats and their attribution to salmonellosis outbreaks. This study then ranked serovars based on this multiplier and designated consistently higher risk serovars across products and time as the higher virulence (HV) serovars. The remaining serovars are referred to hereafter as lower virulence (LV) serovars. To obtain a genetic fingerprint for HV serovars, researchers then used a supervised random forest, an agglomerative machine learning approach, to determine which individual VFs and VF categories are most strongly associated with these HV serovars. To accomplish objective 2, researchers evaluated temporal trends of virulence between 2016 – 2023. Specifically, researchers tested trends in the prevalence of HV serovar and for VF carriage (i.e. the number of unique VFs carried per NTS isolate) in these serovars as well as 15 LV serovars, but that are commonly retrieved from beef, chicken, or from human cases.
64,284 NTS isolates from beef, chicken, and human cases and 33,007 VFs were included in this analysis. Of those 33,007 VFs, 350 were found to be informative in distinguishing between HV and LV serovars. This research identified five HV serovars of relevance in meat products: S. Enteritidis, S. Braenderup, S. Newport, S. Monophasic, and S. Typhimurium, of which the latter three are particularly important to beef given their higher prevalence in that product. Researchers were able to identify a VF profile that predicted HV serovars with over 98% accuracy. In addition, individual VFs, as well as categories of VFs, that were more strongly associated with virulence (e.g., exotoxins) were identified. Consistent with the literature, researchers found that the HV and LV serovar prevalence in beef and in chicken products has remained stable over the 2020-2023 period compared to 2016-2020. © 2024 EpiX Analytics LLC. In contrast, the last four years of data (2020-2023) saw a small but significant increase of 4% (95% CrI: 4.1 - 4.4%) in VF carriage in all isolates, and an increase in 2023 alone of 9% (8.8 - 9.4%). Nonetheless, increases varied across serovars. Certain serovars, like S. Infantis (LV), experienced increases in VF carriages, whilst others, like S. Braenderup (HV), saw minimal genetic changes over the investigated timeframe. Across both HV and LV serovars, certain VF categories are increasing, like nutrition, which increased by 6.6% (6.3 – 6.9%) over the last four years. Others, like exotoxins, decreased by 1.8% (1.2 - 2.3%) over the last four years. Whether this is due to differential fitness advantages conferred by those genes, shifts in NTS population structure, or other factors require further investigation. The changes observed in HV and LV serotypes indicate the need for the inclusion of temporally informed genetic analysis for the prediction of serovars of increasing virulence. This study identified three tentative criteria which can help guide these approaches: 1) Total VF carriage; 2) Carriage of VFs strongly associated with HV genetic profiles, and 3) Carriage of VFs from VF categories of high concern. Each criterion is evaluated with two metrics: 1) proximity to HV values, and 2) temporal changes.
A shift towards focusing on subpopulations (e.g. serovars) of NTS and other pathogens, and an emphasis on virulence genes to implement more risk-based control targets could be followed. As industry explores initiatives to further improve food safety, this project highlights the importance of understanding temporal dynamics of the serovars and candidate genetic targets that industry could use for monitoring and control programs using modern and rapid diagnostic methods. While the prevalence of HV serovars in beef products has remained stable over the last eight years, the presence of some types of genes most associated with virulence has increased in recent years, in both HV and LV serovar groups. This research is a key step towards providing industry with tools to proactively detect and address emerging NTS serovars and genes before they become serious public health threats.