Monitoring the microbiome for food safety and quality using deep shotgun sequencing

This research explores the use of microbiome sequencing to enhance food safety and quality. Current methods like culturing or targeted whole-genome sequencing (WGS) have limitations in fully characterizing complex microbial communities and their interactions within food. Next-generation sequencing (NGS) offers a promising alternative, but requires further development for accuracy, speed, and supply chain applicability. Previous studies have utilized microbiome analysis for understanding flavor and quality in fermented foods, agricultural processes, and manufacturing processes, but its potential for food safety is largely unexplored. While WGS is increasingly used for pathogen detection and outbreak investigation, its reliance on culturing introduces bias. High-throughput sequencing of total RNA, in contrast, offers a culture-independent approach and may provide insights into the viability and active components of the microbial community. Total RNA sequencing has advantages over total DNA or amplicon sequencing in terms of accuracy, reproducibility, and minimizing PCR amplification bias. This study aims to demonstrate the utility of total RNA sequencing for characterizing food microbiomes and identifying potential safety and quality indicators. The researchers focused on high-protein powder (HPP) samples, a common pet food ingredient susceptible to microbial growth, to evaluate their hypothesis that shifts in microbial communities can serve as indicators of food quality and safety.

The introduction section of the paper reviews existing literature on the use of microbiome analysis in food science. Several studies are cited that have used microbiome sequencing to examine aspects of flavor and quality in fermented foods, as well as agricultural and manufacturing processes. The authors point out that while the use of whole-genome sequencing (WGS) for pathogen detection is gaining traction, it has limitations due to its reliance on culturing. The review highlights the potential of high-throughput sequencing of total RNA as a culture-independent method for characterizing microbial communities, emphasizing its advantages over other methods such as DNA sequencing or amplicon sequencing. The need for additional research to validate the use of microbiome sequencing for food safety and quality applications is also mentioned.

The researchers collected 31 high-protein powder (HPP) samples from two suppliers over 1.5 years. Each sample underwent deep total RNA sequencing (~300 million reads per sample). A novel bioinformatic pipeline was developed to process the data. This pipeline incorporated a crucial step for filtering out eukaryotic food matrix RNA, which significantly improved the specificity of microbial identification. Kraken, a k-mer classification tool, was used for this purpose, with a custom reference database created from 31 common food ingredient and contaminant genomes. The effectiveness of the RNA-based taxonomic identification was validated against DNA sequencing results, showing a strong correlation (R² = 0.93). The pipeline further quantified microbial relative abundance using reads per million (RPM), with a threshold of RPM > 0.1 indicating presence. Alpha and beta diversity analyses were conducted to assess microbial diversity within and between samples. To study the relationship between microbiome composition and food matrix composition, the researchers examined samples with unexpected pork and beef content. The ability of total RNA sequencing to identify foodborne pathogens was evaluated by focusing on 14 pathogen-containing genera. Finally, the concordance between *Salmonella* culturability (culture-based testing) and *Salmonella* relative abundance from sequencing was assessed using multiple analysis methods, including alignment to an expanded *Salmonella* genome database and analysis of the elongation factor Tu (ef-Tu) gene.

The developed bioinformatic pipeline, incorporating a crucial eukaryotic matrix filtering step, achieved >99.96% specificity in microbial identification. An average of 119 microbial genera were identified per HPP sample, with 65 genera common to all samples. *Bacteroides*, *Clostridium*, *Lactococcus*, *Aeromonas*, and *Citrobacter* were the most abundant genera. Significant differences in microbial composition were observed between samples from the two suppliers. Samples with unexpected pork and beef contamination showed distinct microbiome shifts, indicating that microbiome analysis can detect adulteration. While several foodborne pathogen-containing genera were detected in all samples, there was no consistent correlation between *Salmonella* read abundance from total RNA sequencing and *Salmonella* culturability from culture-based tests. Although using a more extensive *Salmonella* genome database for alignment improved the concordance, it didn't achieve complete agreement. Co-occurrence analysis revealed associations between *Salmonella* growth and specific genera like *Erysipelothrix*, *Lactobacillus*, *Anaerococcus*, *Brachyspira*, and *Jeotgalibaca*, as well as a negative correlation with *Gyrovirus*. The study highlighted that 2-4% of reads remained unclassified, potentially representing novel microbes.

The study demonstrates the potential of total RNA sequencing for monitoring food microbiomes and detecting safety and quality issues. The improved specificity of microbial identification, achieved through the eukaryotic matrix filtering step, is crucial for accurate microbiome characterization. The observed microbiome shifts associated with food matrix contamination highlight the potential of microbiome analysis as an indicator of adulteration or unexpected ingredients. However, the lack of perfect concordance between *Salmonella* RNA sequencing data and *Salmonella* culturability underscores a limitation of relying solely on RNA sequencing for pathogen viability prediction. This discrepancy could be due to factors such as the presence of nonculturable *Salmonella* or limitations in the reference databases. The study emphasizes the need for broader and more complete reference databases for improved accuracy and the importance of integrating multiple data sources (e.g., microbiome composition, culture-based results, and microbial load) for a comprehensive assessment of food safety.

This study successfully developed a robust bioinformatic pipeline for analyzing food microbiomes using deep total RNA sequencing. The pipeline's ability to identify unexpected matrix contaminants in HPP samples demonstrates its potential for food safety applications. While RNA sequencing is valuable for characterizing complex food microbial communities, additional research is necessary to improve prediction of pathogen viability. Future research directions include expanding microbial reference databases, exploring the use of RNA sequencing for detecting antimicrobial resistance and virulence factors, and further benchmarking against culture-based methods.

The study's main limitation is the lack of perfect correlation between *Salmonella* abundance detected by RNA sequencing and *Salmonella* culturability. This could be due to limitations in current *Salmonella* reference databases, the presence of non-culturable *Salmonella*, or other factors influencing *Salmonella* viability. Another limitation is the sample size; a larger and more diverse sample set would strengthen the generalizability of the findings. The study focused on a specific food ingredient (HPP) and the findings might not be directly transferable to other food types.