Source: CORNELL UNIVERSITY submitted to
DEFINING NICHE-SPECIFIC MICROBIAL COMMUNITY DYNAMICS TO INFORM TARGETED SANITATION AND HYGIENIC DESIGN
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
ACTIVE
Funding Source
Reporting Frequency
Annual
Accession No.
1027904
Grant No.
2022-67017-36289
Cumulative Award Amt.
$625,100.00
Proposal No.
2021-08139
Multistate No.
(N/A)
Project Start Date
Jan 1, 2022
Project End Date
Dec 31, 2025
Grant Year
2022
Program Code
[A1332]- Food Safety and Defense
Project Director
Snyder, A.
Recipient Organization
CORNELL UNIVERSITY
(N/A)
ITHACA,NY 14853
Performing Department
Food Science
Non Technical Summary
Our goal is to prevent pathogens from getting into the food supply. Food production in the U.S. often takes place in large, industrial plants with complex niches (i.e. places where pathogens can "hide") that resist sanitation efforts. However, little analysis has been done on the microbial communities living in these niches, including which attributes of the niche drive the selection for particular microbes. Most environmental sanitation research has not studied the role of sanitizer fluid flow in and out of the niche even though we know that it greatly impacts microbial removal. We aim to resolve these unknowns by identifying and defining specific microbial communities within different niches to better understand the sources of important foodborne pathogens likeListeria monocytogenes. We will collect more than 1,000 environmental samples from three large dairy plants to define the microbiome of different niches. We will then apply machine learning algorithms to identify the features most predictive of specific microbial taxa, verifying these associations using laboratory studies. Finally, we will explore how complex features of the niche impact environmental sanitation by restricting fluid flow, using sanitation challenge studies to determine how sanitation impacts diverse microbial communities. Overall, our work aims to reduce outbreaks and recalls; advance source tracking and root cause analysis after cross-contamination; and support better design of food plants to strategically eliminate microbial harborage points.
Animal Health Component
50%
Research Effort Categories
Basic
50%
Applied
50%
Developmental
(N/A)
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
71234301103100%
Goals / Objectives
Thelong-term goalof this proposal is to reduce environmental cross-contamination during food manufacturing. A barrier to achieving this goal is thatcurrent environmental sanitation strategies often neglect niche-specific differences in microbial load and community composition. We propose leveraging a combination of metagenomic, machine learning, and Multiphysics modeling tools to define the attributes of niches which drive microbial selection and mediate the physical aspects of sanitation. Our findings will guide the development of targeted principles in sanitation and hygienic design that will reduce the incidence of pathogen cross-contamination that leads to outbreaks and recalls.We aim to define the specific microbial communities within different environmental niches to better understand the reservoirs of environmental pathogens likeL. monocytogenes. We will collect 1,200 environmental samples from three large dairy plants and use targeted metagenomics to define the microbiome of different niches. We will then apply machine learning algorithms to identify the features most predictive of specific microbial taxa, verifying these associations using laboratory studies. Finally, we will explore how complex features of the niche impact environmental sanitation by restricting fluid flow, quantifying wall shear stress and using sanitation challenge studies to determine how sanitation impacts diverse microbial communities.Overall goals:Identifying the locations in food plants that harbor particular bacterial and fungal populations will improve our ability to assess risk and more efficiently track and eliminate contaminants. The findings from this work will enable plant staff to more rapidlyidentify the most effective interventions, rather than having to try a range of potential interventions in the hopes of addressing the root cause. Targeted interventions will in turn save the plant money. Our work can also enable more precise assessments ofL. monocytogenescross-protection within biofilms so that plants can more strategically apply enhanced sanitation. Overall, we aim to reduce the incidence of cross-contamination, decrease the likelihood and duration of outbreaks, reduce food waste, limit the application of excessive and wasteful sanitizer, and increase customer satisfaction.Structural design is strongly predictive of pathogen harborage. However, capital investments are expensive, and it is difficult to ensure that niches will be eliminated. Additionally, preventative maintenance, such as the regular replacement of wearable goods, has been shown to improve control over environmental biota. Yet, with limited direct evidence, maintenance staff are often left to self-determine the frequency and nature of these activities. In the long term, our work can help plants identify problematic niches by their attributes and implement strategic capital improvements in hygienic design and preventative maintenance.
Project Methods
We will use metagenomics, classical microbiology, and microscopy to establish signature patterns in surface-associated microbiota harbored in common types of environmental niches.Establishing environmental reservoirs for different microbial communities will support source-tracking investigations in the event of cross-contamination.At each plant, 50 environmental niches will be identified for repeated sampling. These 50 sites will be comprised of 10 different types of niches distributed across equipment surfaces, architectural features of the building, and utensils. The PI will participate in sample collection at each site to ensure consistent identification of swab sites. For classical microbiological analysis, samples (100µl, 100to 10-6dilutions) will be spread plated on tryptic soy agar (TSA) and on malt extract agar (MEA). Replicate plates will be incubated at 10°C and 30°C to address differences in the surface temperature of the niche from which they were collected. TSA plates at 30°C will be incubated for 48 hr while TSA plates at 10°C will be incubated for at least 7 d. All MEA plates will be incubated for 7 d to accommodate slower growing fungi. Filamentous fungi (molds) and yeasts will each be enumerated separately (51, 126). Additional plating will be performed on crystal violet tetrazolium agar (CVTA), polymyxin pyruvate egg-yolk mannitol-bromothymol blue agar (PEMBA), and MRS agar to obtain counts of viable cells within broad categories of problematic environmental biota. The remainder of the re-suspended cell pellet (~800µl) will be immediately frozen at -80°C. Comparisons among microbial counts from different selective media will be evaluated using a mixed model ANOVA to account for multiple plants and sampling dates. The model will be fitted using the lme4 package in R. Significant differences in viable cell counts among niche types will be evaluated using the emmeans package. For metagenomic data, principle coordinate analysis (PCoA) will be carried out using weighted UniFrac distances to visualize the ordination and potential associations among samples as part of an initial exploratory analysis. Pairwise permutational multivariate ANOVA (PERMANOVA) will be used with the UniFrac distance matrix to identify potential associations between microbial community composition and niche type using the pairwise adonis function.We will identify the characterizing features of niches highly associated with specific microbial communities using random forest analysis. We will probe beyond simple categories of niche types into the empirical properties of niches. We will use advanced data analytics in hypothesis generation and test those hypothesized causal relationships using controlled laboratory studies of model niches. We will add point specific non-group variables (e.g. niche attributes) to the PERMANOVA analysis using the adonis function in R. This exploratory analysis will help identify which taxa are associated to particular niche attributes. Associations between each pair of samples will be determined using the UniFrac similarity measure. Patterns will be visualized using the PCoA and superimposing symbols to represent the factor of interest onto the PCoA plots. The effect of the point specific non-group variables on the niche-associated taxa will be quantitatively tested using PERRMANOVA. This will also enable us to explore differences based on facility and sampling date.Next, a supervised machine learning algorithm for multivariate random forest analysis willbe applied to evaluate the threshold of different niche attributes that are most associated with detection of different microbial taxa. Multivariate random forest will be conducted using the Integrated Multivariate Random Forest package in R (143). A consensus regression tree will be produced for visualization. Variable importance scores will be calculated from the multivariate random forest analysis and compared among niche features. The top 10 features with the highest variable importance score will be visualized using tornado plots.Additionally, we will explore at least two different approaches to machine learning and compare outcomes, benefits, and drawbacks from these different approaches. Collaboration with Dr. Erika Mudrak of the Cornell Statistical Consulting Unit will be implemented to ensure the appropriateness of method design for quantitative analyses, as described in our letters of support.We will use computational fluid dynamics (CFD) to quantify shear stress during sanitation in niches of known dimension and configuration. We will determine changes to the microbiome within these niches in laboratory sanitation studies. Additionally, we will compare changes in microbial community obtained under laboratory studies to changes observed during in-plant sample collection before and after sanitation shifts.We will select 3 different niche types (e.g. a section of conveyer belt, cart wheel, drain grate) in CFD simulations and sanitation challenge studies. The goal is to develop a realistic assessment of shear stress in real-world niches and determine the consequent microbial outcomes following sanitation. We will compare how well the simplistic models in Experiment 2 approximate the outcomes from the more complex niches tested in Experiment 1. The computational software COMSOL Multiphysics will be used to solve for surface shear stress as the major extraction of non-measurable parameters under the simulated flow conditions. Shear stress values will be extracted using wall functions in the CFD module of COMSOL Multiphysics. Shear stress values will be determined based on the velocity of water that the sanitation sprayer units described below are capable of delivering. Mesh creation will be performed and a computational grid will be created to describe the controlled volume occupied by the fluid inside the niche. Velocity and wall shear stress will be determined by the computational model to be stored in the cells. The geometry of the niche will be created based on the actual dimensions, location, and shape of the fabricated niche. A set of discrete algebraic equations for flow variable simulation will be obtained using a control-volume approach and integrating the governing equations over each cell in the mesh. The flux of fluid through the cell faces will be obtained through interpolation using different numerical techniques so that all the fluid variables are found at each cell node. The flow will follow a laminar model and the velocity magnitude contour across the surface will be obtained from the convergence of the simulation.We will consider more realistic multi-phase flow conditions in our experimental design.To determine the effect of sanitation in these niches on microbial reduction, standardized sanitation challenge studies will be used for both the simplistic model niches and the complex niches.Sanitation parameters will be controlled across a range of application levels including the spray flow velocity, temperature, distance of the nozzle from the surface, application angle, sanitizer chemistry, concentration, and contact time. Viable cells will be enumerated by replicate plating on general purpose and selective-differential media (dependent on the inoculum) so that survivorship can be compared among broad categories of cell types.

Progress 01/01/23 to 12/31/23

Outputs
Target Audience:Cross-contamination in food manufacturing plants is a significant challenge, particularly for ready-to-eat (RTE) foods. The salience of this topic can be seen in the incidence of outbreaks due to post-processing contamination; the application of molecular subtyping, which has revealed the persistence of pathogens in the manufacturing environment; and the introduction of sanitation as a preventive control within the Human Food Rule of FSMA (21 CFR 117.135). But while advancements in detection and subtyping methods have further elucidated the issue, evidence-based interventions remain underdeveloped.In processing facilities where product is not contained within packaging following the kill-step, or where no kill-step is applied, the risk of cross-contamination from persistent colonizers of the processing environment is a significant threat to food safety.This challenge cuts across food sectors and includes dairy (e.g., cheeses, ice cream), meat (e.g., deli products), fresh produce (e.g., leafy greens, fresh-cut melon), and recipe meals (e.g., frozen dinners, assembly meals, RTE sandwiches and salads).In this project, we selected dairy processing plants as model food processing environments because of their highly industrialized processing equipment and many niches, consistent processing and sanitation cycles, and well characterized product microbiological standards.However, we want to also acknowledge that many other food sectors are also served by addressing this challenge. In addition to the dairy industry, the fresh produce industry represents other important stakeholders for this work.We do not envision that the outcomes from this project nor the conceptual framework of the project are limited only to dairy processing environments. While further future work would be needed to define the identity of members of the microbiome within other operation types, we believe that integrating leaders from both the dairy and the fresh produce industry within our discussions can help drive the expansion of this work across product sector categories. Therefore, we have identified industry advisors from both fields as well as a technology company to support technology transfer to diverse audiences. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?Collaborative Opportunities This project has provided a unique opportunity for collaboration in our research group.We have a team consisting of two postdocs, three Ph.D. students,a master's student, and an undergraduate researcher that have assisted in sample collection, processing, and analysis.Each member of our team has different expertise and background which has provided in-depth insight into the various aspects of project design, execution, and analysis.Bi-weekly meetings have been held since the beginning of the project in which issues, strategies, and suggestions for sample collection have been discussed.Meetings also include the development of individual research projects which will use this data set.We have also collaborated with a faculty members from other institutions. Industry engagement for trainees A team including the PI, two Ph.D. students and a postdoc have attended the sample collection trips.During these trips, everyone has had the opportunity to engage with the industry professionals at the facilities to learn about facility management, operation, and challenges. Publication Opportunities The dataset from this project will be used in at least 4 different research papers which will be published in peer-reviewed journals and presented at research conferences.We have also had the opportunity to present our work at weekly lab-group meetings in the Department of Food Science at Cornell. Mentorship This project has provided several opportunities for teaching and mentorship for a postdoc and graduate students.The postdoc has taken on a significant role mentoring a master's student who is analyzing the data from this project for her thesis as well as leading and organizing biweekly meetings for the research team. How have the results been disseminated to communities of interest? The code from this project will be publicly available on Github. We've shared updates about our microbiology results with our industry partners. The first manuscript from our paper will be submitted Spring 2024. The 16S sequences collected from our project have been made publicly available on NCBI Project NumberPRJNA1075764https://www.ncbi.nlm.nih.gov/sra/PRJNA1075764 What do you plan to do during the next reporting period to accomplish the goals? We are currently working on two projects that will investigate how food surface microbiomes (bacterial and fungal) are different between facility environments and seasons.This project will illuminate how the processing environment can influence the taxonomic and ecological characteristics of food surface microbiomes.We plan to collect another batch of environmental samples from a facility of interest to investigate for this project. We plan to develop a project to use the metadata we collected to predict individual taxa on different food surfaces. We plan to develop guidelines for effective feature selection of the data attributes to better characterize and predict environmental surface microbiota. Another project will address the third objective from the proposal and use computational fluid dynamics to model sanitation fluid flow through the environmental niches.

Impacts
What was accomplished under these goals? Sample collection An additional 163 samples were collected in Spring 2024 from three different facilities. We collected samples from over 25 different environmental niches. We gathered information on the spatial dimensions of the niches.These samples were processed using classical microbiology techniques to identifyListeria, spore formers, yeast & molds, Enterobacteriaceae, and Gram-negative bacteria. Sequencing We finished the ITS and 16S sequencing data from all of our samples.The Qiagen DNeasy PowerSoil DNA extraction kit was used to extract the DNA from the samples.A portion of the samples had too low of a bioload for DNA extraction and were removed from analysis.The ITS2 region and the 16S V4 regions were selected for sequencing and sent to Novogene.The results were processed in QIIME 2.A protocol for the appropriate analysis of these sequences was selected based on the attributes of the samples.Then, the taxonomy was assigned using the GreenGenes database for 16S sequences and the UNITE database for the ITS sequences.The alpha and beta diversity were calculated.Visualizations of phylogenic trees, relative abundance, microbial network maps, and diversity were developed in Rstudio. Manuscript Preparation A manuscript lead by a postdoc to assess the discrepancies among diversity and taxonomic characterizations generated from eight different bioinformatic pipelines analyzing the 16S RNA sequences collected from the dairy processing facility surfaces has been developed.The manuscript will be submitted in Spring 2024. Summary of Findings: Accurate knowledge of the microbiota on food processing surfaces is important for food quality and safety.However, little has been done to assess the impact of bioinformatic pipeline selection on the accuracy of food processing surface microbiota characterizations.Our research assessed discrepancies among diversity and taxonomic characterizations generated from eight different bioinformatic pipelines analyzing 16S RNA sequences from dairy processing facility surfaces.We found that characterization of low abundance genera (i.e.,below 1% relative abundance) and diversity outcomes were significantly impacted by bioinformatic pipeline selection.Specifically, the choice of sequence aggregation and normalization methods had the most influence on these outcomes.Pipelines using centered log-ratio (CLR)-transformation inflated diversity metrics compared to pipelines using rarefaction, suggesting that rarefaction is preferable for retaining accurate characterization of uneven microbiomes.Furthermore, there were significant discrepancies in diversity values and genera characterizations between the amplicon sequence variant (ASV) and operational taxonomic unit (OTU) approaches.Pipeline selection requires close consideration of microbiota structure to minimize bias.Increased investigation of bioinformatic algorithms are required for accurate ecological interpretations from 16S rRNA sequences.

Publications

  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Chen, H, Moraru, C (2023). Exposure to 222 nm far UV-C effectively inactivates planktonic foodborne pathogens and inhibits biofilm formation. Innovative Food Science & Emerging Technologies. 103411.


Progress 01/01/22 to 12/31/22

Outputs
Target Audience:Cross-contamination in food manufacturing plants is a significant challenge, particularly for ready-to-eat (RTE) foods. The salience of this topic can be seen in the incidence of outbreaks due to post-processing contamination; the application of molecular subtyping, which has revealed the persistence of pathogens in the manufacturing environment; and the introduction of sanitation as a preventive control within the Human Food Rule of FSMA (21 CFR 117.135). But while advancements in detection and subtyping methods have further elucidated the issue, evidence-based interventions remain underdeveloped.In processing facilities where product is not contained within packaging following the kill-step, or where no kill-step is applied, the risk of cross-contamination from persistent colonizers of the processing environment is a significant threat to food safety.This challenge cuts across food sectors and includes dairy (e.g., cheeses, ice cream), meat (e.g., deli products), fresh produce (e.g., leafy greens, fresh-cut melon), and recipe meals (e.g., frozen dinners, assembly meals, RTE sandwiches and salads).In this project, we selected dairy processing plants as model food processing environments because of their highly industrialized processing equipment and many niches, consistent processing and sanitation cycles, and well characterized product microbiological standards.However, we want to also acknowledge that many other food sectors are also served by addressing this challenge. In addition to the dairy industry, the fresh produce industry represents other important stakeholders for this work.We do not envision that the outcomes from this project nor the conceptual framework of the project are limited only to dairy processing environments. While further future work would be needed to define the identity of members of the microbiome within other operation types, we believe that integrating leaders from both the dairy and the fresh produce industry within our discussions can help drive the expansion of this work across product sector categories. Therefore, we have identified industry advisors from both fields as well as a technology company to support technology transfer to diverse audiences. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?Bi-weekly team meetings This project has provided a unique opportunity for collaboration in our research group.We have a team consisting of two postdocs, three Ph.D. students, a professional master's student, and several undergraduates that have assisted in sample collection, processing, and analysis.Each member of our team has different expertise and background which has provided in-depth insight into the various aspects of project design, execution, and analysis.Bi-weekly meetings have been held since the beginning of the project in which issues, strategies, and suggestions for sample collection have been discussed.Meetings also include the development of individual research projects which will use this data set. Industry engagement for trainees A team including the PI, two PH.D. students and a postdoc have attended the sample collection trips.During these trips, everyone has had the opportunity to engage with the industry professionals at the facilities to learn about facility management, operation, and challenges. PD opportunities with teaching and presenting The dataset from this project will be used in at least 4 different research papers which will be published in peer-reviewed journals and presented at research conferences.We have also had the opportunity to present our work at weekly lab-group meetings in the Department of Food Science at Cornell. How have the results been disseminated to communities of interest?To date, results have primarily been shared through (1) results reporting to collaborating industry stakeholders and (2) through initial publication of research papers. In the future, as we generate additional results and more complete analyses, we anticipate sharing more broadly through communities of interest. Currently, we have provided our collaborating sites with their results forListeriatesting and with a summary of yeast and mold findings. We have published one paper based on our preliminary work establishing a metadata scheme, which was used by collaborators at FDA in the development of a metadata standard currently available through NCBI. Our team also has a second manuscript under review, led by co-PI Moraru. What do you plan to do during the next reporting period to accomplish the goals?We have one more sampling time point among the three sites.All the data from all sampling seasons will then be combined, processed, and analyze.Specifically, we will be conducting an in-depth analysis examining the effect of the bioinformatic pipeline on amplicon sequencing results as one of our first products from this data set.The samples we collected ranged from low-density, low-diversity samples to high-density, high-diversity samples and we hypothesize that there will be significant differences between diversity results based on step selection within the pipeline.Our team is also examining methods to improve DNA extraction from spore formers.Lysing spores can be inefficient or incomplete during DNA extraction.Since our classical microbiology results have detected spore formers in many samples, this project is crucial for ensuring our results sufficiently encompass the diversity of our samples.Both projects will have significant implications for environmental sampling and characterization.

Impacts
What was accomplished under these goals? Sampling exercises have been completed for ¾ of a year We have collected474environmental samples from three different dairy plants in New York State.Each dairy plant has been visited3 times(Summer 2022, Fall, 2022, and Winter 2023).One more set of visits is planned for Spring 2023 where we plan to collect an additional175 samples.We obtained samples fromover25different environmental niches.Furthermore, these samples were processed using classical microbiology techniques to identifyListeriaand quantify spore formers, yeast & molds, Enterobacteriaceae, and Gram-negative bacteria. Sequencing results The Qiagen DNeasy PowerSoil DNA extraction kit was used to extract the DNA from the Summer 2022 environmental samples.A portion of the samples had too low of a bioload for DNA extraction and were removed from analysis.The ITS2 region and the 16S V4 regions were selected for sequencing.The results were processed in QIIME 2.A protocol for the appropriate analysis of these sequences was selected based on the attributes of the samples.Then, the taxonomy was assigned using the GreenGenes database for 16S sequences and the UNITE database for the ITS sequences.The alpha and beta diversity were calculated.Visualizations of phylogenic trees, relative abundance, microbial network maps, and diversity were developed in Rstudio. Metadata collection method is publicly available We developed a detailed metadata standard which we used to characterize swab site descriptions during collection, which we published as part of the preliminary work for this project:https://journals.asm.org/doi/abs/10.1128/msystems.01284-22. This schema was used in the collection of sample metadata for the dairy samples collected in this study. We collected information of niche-specific attributes with15 features collected(ATP, surface roughness, moisture, temperature, humidity, object type, structure, material, angle, risk zone, location in facility, shape, surface continuity, sanitation regime, and soil).After every sampling trip the metadata was cleaned and attached to sample identifiers in Microsoft Excel for further analysis.The metadata files are available to all members of our research team.

Publications

  • Type: Journal Articles Status: Under Review Year Published: 2023 Citation: Chen, H, Moraru, C (2023). Exposure to 222 nm far UV-C effectively inactivates planktonic foodborne pathogens and inhibits biofilm formation. In Review.
  • Type: Journal Articles Status: Published Year Published: 2023 Citation: J. Feng, D. Daeschel, D. Dooley, E. Griffiths, M. Allard, R. Timme, J.B. Pettengill, Y. Chen, A. B. Snyder. 2023. A schema for digitized surface swab site metadata in open-source DNA sequence databases. mSystems. E01284-22.