Recipient Organization
KANSAS STATE UNIV
(N/A)
MANHATTAN,KS 66506
Performing Department
(N/A)
Non Technical Summary
Salmonella remains a persistent food safety challenge, especially in poultry products like ground turkey, which are a significant source of human illness in the United States. Despite many efforts to reduce contamination, Salmonella continues to cause outbreaks, highlighting the need for a better understanding of how and where it appears during turkey production and processing. Traditionally, food safety controls have focused on general practices to reduce contamination across the board. However, recent evidence suggests that these efforts may overlook key differences between how Salmonella survives on the surface (external contamination) versus inside the bird (internal contamination). This project aims to address this critical gap in understanding by investigating Salmonella contamination at the level of individual turkey carcasses as they move through processing.While regulatory agencies like the USDA's Food Safety and Inspection Service (FSIS) have made progress in reducing Salmonella levels in poultry, current public data and scientific studies are limited, particularly for turkeys, which differ from more commonly studied broilers. As regulatory expectations increase, the turkey industry lacks detailed, science-based information needed to guide better control strategies, product testing, and food safety decisions. There is also growing interest in understanding "microbial independence," or how contamination in one bird may (or may not) relate to contamination in another during processing. This matters because such information can help producers define lots of product more accurately and use risk-based sampling methods that are both effective and practical.To fill these critical knowledge gaps, this research follows a detailed plan with three major aims. The first aim is to determine how much Salmonella contamination varies from one turkey carcass to another during processing, and to distinguish between external and internal contamination. Researchers will collect samples from the same birds at multiple processing points, from early stages like pre-scald to final products like ground turkey. They will use advanced tools, including rapid detection methods, microbial counts, and deep genetic typing (CRISPR-SeroSeq), to track how Salmonella behaves at each stage. This will reveal which types (serotypes) of Salmonella are most common and whether they pose a public health concern.The second aim is to use the data to build a new framework for understanding how Salmonella populations behave during processing. By analyzing both microbial levels and serotypes, researchers will identify patterns that show how processing steps like washing, chilling, and grinding affect internal and external contamination differently. For example, they will model how much contamination is reduced by surface treatments (such as sanitary dressing) and whether internal contamination re-emerges later in ground product. This modeling also includes statistical tools to detect whether contamination in one bird might be linked to another, helping to determine microbial independence. These insights can guide smarter process controls and lotting strategies, making it easier for processors to identify when a problem starts and how far it might spread.The third aim is to apply this framework in real-world settings to improve how the turkey industry approaches Salmonella control. This includes working directly with a commercial turkey processor to test the framework under production conditions, and organizing collaborative sessions with industry stakeholders. These "co-creation" meetings will ensure that recommendations are not only scientifically sound but also realistic and useful for processors. The ultimate goal is to create data-driven strategies that allow the turkey industry to take a more proactive role in food safety, identifying the most critical points for intervention and targeting efforts where they will have the greatest impact.This work is carried out by a multidisciplinary team with expertise in microbiology, epidemiology, poultry science, data science, and food safety regulation. The team's combined experience ensures that findings will be scientifically rigorous, practically relevant, and tailored to meet the needs of both regulators and industry. By combining laboratory methods, statistical modeling, and industry collaboration, this project aims to transform how we understand and manage Salmonella in turkey production. It will provide key data that can shape smarter regulations, reduce foodborne illness, and make turkey products safer for consumers.
Animal Health Component
70%
Research Effort Categories
Basic
30%
Applied
70%
Developmental
(N/A)
Goals / Objectives
Aim 1. Determine the variability in Salmonella contamination on turkey carcasses throughout the processing continuum and establish differences in internal and external Salmonella populations.Aim 2. Utilize data collected on Salmonella variability to develop a conceptual framework to define microbial independence and determine internal and external Salmonella populations that are differently affected by processing.Aim 3. Apply the conceptual framework to the turkey industry to improve Salmonella control strategies.
Project Methods
This study employed a multi-stage sampling and analytical approach to quantify and characterize Salmonella dynamics in commercial turkey processing. Samples were collected from individual carcasses at multiple processing stages--pre-scald, hot rehang, post-chill, debone, and ground--using validated poultry methods. Initial screening for Salmonella was performed with the Hygiena BAX® Real-Time PCR system. PCR-positive samples were enriched in tetrathionate (TT) broth, incubated overnight at 37°C, and plated on selective agar for confirmation. Enumeration was conducted using high-throughput most probable number (MPN) methodology in 96-well deep-well plates, with MPN values calculated using the GalaxyTrakr MPN calculator. Only PCR-positive samples were quantified.To analyze differences in Salmonella prevalence and levels across processing stages, a mixed-effects model was applied, accounting for repeated measures on individual carcasses. Fixed effects (e.g., processing stage) were selected using fit statistics including AIC, BIC, and PRESS. To better understand contamination patterns, external Salmonella levels (pre-scald to post-chill) were modeled using lognormal distributions, fitted with censored data methods to incorporate both quantifiable and non-detect results. Internal contamination, inferred from ground product samples, was modeled as a mixture of two lognormal distributions: one representing residual external contamination (parameters fixed from post-chill data) and the other representing internal contamination. Fitting was performed in R and/or @RISK software, with best-fit models evaluated using AIC and BIC. This approach allowed estimation of both the internal population parameters (mean, standard deviation) and the probability that a given observation was attributable to internal contamination.Serotyping was conducted via CRISPR-SeroSeq, which uses a single PCR to amplify native CRISPR spacers from overnight TT cultures. Libraries were barcoded and sequenced using Illumina NextSeq, and serotypes were identified via BLAST alignment to a reference database of over 150 Salmonella serotypes. The resulting data were processed in R (v4.4.1) and visualized using heatmaps. Relative serotype frequencies were used to characterize changes across process stages and classify serotypes as abundant, background, or of public health concern (e.g., Hadar, Typhimurium, Muenchen). Serotype distribution modeling paralleled level modeling: external population serotypes were directly observed from pre-scald through post-chill, while internal population serotypes were inferred from ground product profiles. These were modeled as mixtures where internal contributions were estimated based on the divergence from post-chill serotype distributions, adjusted using the previously estimated internal observation probability.To interpret how processing affects Salmonella levels and sources, serotype changes were paired with quantitative level shifts. Observable reductions from pre-scald to post-chill were attributed to external contamination. Conversely, appearance of new serotypes in post-chill, deboned, or ground product--especially when absent earlier--was attributed to internal contamination. Complex patterns, such as persistence of a serotype throughout processing or level increases later in processing, were probabilistically assigned to either population based on context. Once assigned, internal and external population observations were fitted to separate lognormal distributions using censored fitting techniques, supporting precise estimation of population-specific behavior across stages.To quantify microbial independence, autocorrelation analysis was performed on data from 75 sequentially sampled, individually tagged carcasses on a single processing line. This assessed statistical correlations between Salmonella levels in adjacent carcasses at each process stage using lag analysis. Lag 1 (correlation with immediately previous carcass), Lag 2, and so forth were examined to determine if contamination levels clustered or were independently distributed. Statistically significant correlations--especially at shorter lags--indicated potential dependencies, offering foundational data to define microbial independence in carcass-level contamination and informing risk-based lotting and sampling strategies.Finally, repeated measures on individual carcasses were used to model shifts in internal and external population mixtures over time. These population dynamics were analyzed for each processing stage to determine how interventions (e.g., scalding, chilling, grinding) impacted population prevalence, level, and variability. Observed changes in Salmonella levels between stages were categorized as (i) quantifiable reductions (e.g., 5-log to 3-log = 2-log reduction), (ii) censored changes (e.g., quantifiable to