Microbiome-based environmental monitoring of a dairy processing facility highlights the challenges associated with low microbial-load samples

Dairy processing environments harbor microorganisms that can contaminate food products. Traditional monitoring methods, like agar plating, are limited in their scope and accuracy, often resulting in false positives. DNA sequencing offers a more comprehensive approach, enabling identification of the entire microbial population and providing insights into the functional potential of species, including spoilage and virulence properties. However, high-throughput sequencing methods are often expensive and require specialized expertise. Portable sequencing devices, like the ONT MinION, offer a potential solution by simplifying workflows and reducing the need for highly skilled personnel. This research aimed to evaluate the feasibility and accuracy of MinION sequencing for environmental monitoring in a dairy processing facility, comparing its results with those obtained using Illumina sequencing and traditional culture-based methods. The study aimed to determine if MinION-based rapid sequencing could accurately identify and distinguish between related microorganisms commonly found in dairy processing environments, and then compare the results obtained using MinION with those from Illumina sequencing and culture-based methods in a real-world setting of environmental swabs from a food processing facility.

Previous studies have demonstrated the use of DNA sequencing to characterize dairy and environmental samples, enabling source tracking and improved understanding of microbial population composition. High-throughput metagenomic sequencing provides superior insights compared to culture-based methods, uncovering information on species' functional potential. However, traditional high-throughput approaches often require significant resources, limiting their routine implementation in food processing settings. While portable sequencing technologies, such as the ONT MinION, have shown promise in clinical settings for rapid pathogen identification, their application in food processing environments for environmental monitoring remained largely unexplored before this study.

The study involved two phases. First, a proof-of-concept experiment using a mock community of four related spore-forming bacteria (*Bacillus cereus*, *Bacillus thuringiensis*, *Bacillus licheniformis*, and *Geobacillus stearothermophilus*) was conducted. Both 16S rRNA gene amplicon sequencing and whole metagenome sequencing (WMGS) were performed using the ONT MinION. The second phase involved environmental monitoring of a dairy processing facility. Eight locations were swabbed on three different days across three months. Swabs were processed for MinION sequencing (using MDA for DNA amplification) and Illumina NextSeq sequencing (both with and without MDA). Additionally, a culture-based analysis was performed, including plating on BHI agar and subsequent 16S rRNA Sanger sequencing of selected colonies. Bioinformatic analysis involved various tools for sequence alignment, taxonomic classification (MEGAN, Kraken2, Bracken, MetaPhlAn2), genome assembly (Canu, OPERA-MS), and metagenome-assembled genome (MAG) binning (MetaBAT2). Statistical analyses, including Pairwise Wilcoxon rank sums tests with Benjamini Hochberg correction, were performed to compare the results obtained using different methods and sample preparation techniques.

MinION sequencing accurately identified species in the mock community, demonstrating its capability to distinguish between closely related microorganisms. In the environmental monitoring phase, both MinION and NextSeq sequencing provided comparable species-level taxonomic classifications, with some discrepancies. The most abundant species identified included *Kocuria* sp. WRN011, *Enterococcus casseliflavus*, and *Enterococcus faecium*. There was some overlap between species identified in environmental samples and negative controls, particularly in low biomass areas, highlighting the risk of false positives. Metagenome-assembled genome (MAG) analysis revealed high-quality genomes from environmental isolates and positive controls. MDA amplification introduced bias, affecting the relative abundances of certain species. Culture-based analyses showed lower diversity than culture-independent methods, indicating selection bias. Overall, only six out of 108 species showed significant differences in relative abundance based on processing or sequencing method. Twenty-four out of 46 genera exhibited significant differences based on method used. Several genera showed significant differences across multiple pairwise comparisons, including *Pseudochrobactrum*, *Exiguobacterium*, and *Planococcus*.

The study demonstrated that MinION sequencing offers a viable alternative to Illumina sequencing for microbiome analysis in food processing environments, providing comparable accuracy at species level. However, the requirement for high DNA concentration remains a challenge, particularly in low-biomass samples where MDA amplification is necessary and may introduce bias. The high degree of similarity between negative controls and low-biomass samples highlights the importance of careful control design and stringent data interpretation. The generation of MAGs from this study greatly enhances the understanding of the environmental isolates and provides resources for future investigations into food processing microbial ecology. The observed differences in relative abundances due to different processing and sequencing methods emphasize the need for standardization and careful consideration of these factors when interpreting microbiome data.

This study demonstrates the potential of MinION sequencing for rapid microbiome analysis in food processing settings, while highlighting challenges related to low DNA yields and potential biases introduced by amplification and culturing. The generation of high-quality MAGs from environmental isolates offers a valuable resource for future research. Future research should focus on improving DNA extraction methods for low biomass samples, further optimizing MinION protocols, and developing standardized bioinformatics pipelines for metagenomic data analysis in food chain applications. The development of portable technologies that require less DNA input would greatly benefit this field.

The study was limited to a single dairy processing facility, and the generalizability of findings to other facilities may be limited. The use of MDA amplification could have introduced bias into the results. The relatively low sequencing depth might have influenced the detection of less abundant microbial taxa. The relatively small sample size (8 sampling locations) and short duration of study may not fully capture the dynamic nature of microbial communities in food processing environments.