Skip to main content
ARS Home » Northeast Area » Wyndmoor, Pennsylvania » Eastern Regional Research Center » Molecular Characterization of Foodborne Pathogens Research » Research » Research Project #429773

Research Project: Advanced Development of Innovative Technologies and Systematic Approaches to Foodborne Hazard Detection and Characterization for Improving Food Safety

Location: Molecular Characterization of Foodborne Pathogens Research

2021 Annual Report

1: Development and evaluation of technologies for sample preparation, detection (label-based and label-free), and characterization of microbial, chemical, and biological contaminants of concern in foods that can be implemented for improved food safety and food defense. 1A. Spectroscopy-based identification of foodborne pathogens, toxins, and chemical contaminants. 1B. Antibody-based detection of foodborne pathogens and toxins. 1C. DNA - based detection of foodborne pathogens. 1D. Phage-based detection of foodborne pathogens. 1E. Cell-based detection of foodborne pathogens and toxins. 1F. Enabling technologies for pathogen detection. 2: Application of CFSE developed technologies either alone or in combination with existing methods to evaluate microbial populations and the microbial ecology of foods during production and processing. 2A. Expand the databases for BARDOT and HESPI techniques using pure cultures of known microorganisms (foodborne pathogens and indicator microorganisms) 2B. Apply BARDOT and HESPI techniques to analyze microbial populations in foods

The food supply must be protected from pathogens, toxins, and chemical contamination that cause disease or illness in humans. Detection technologies are a critical component for identifying and controlling the potentially harmful food contaminants. The overarching goal of the Center for Food Safety Engineering (CFSE), working in collaboration with USDA-ARS scientists, is to develop, validate, and implement new technologies and systematic approaches for improving food safety. We propose to develop a variety of timely, accurate, and cost-effective technologies for the pre-screening, detection, characterization, and classification of foodborne hazards. Our prototype pre-screening and detection technologies include hyperspectral light scattering, metal-enhanced plasma spectroscopy, phage-based detectors, cell-based assays, antibody- and DNA-probe inkjet-printed test strips, plasmonic ELISA, and enhanced lateral flow immunosensors. The accompanying algorithms and software for data processing, analysis, and interpretation of colorimetric, fluorometric, light-intensity, light-scattering, and spectroscopy-based assays, along with time-temperature tracking devices, will enable and enhance these technologies. These methods will detect Listeria monocytogenes, Shiga toxin-producing Escherichia coli (STEC), Campylobacter jejuni, and Salmonella enterica serovars, with demonstrated applications in meat, poultry, and produce, as well as detect toxins, metals, and chemicals of concern in foods. An experienced multidisciplinary team of investigators from Purdue University, the University of Illinois, and USDA will produce and evaluate operational technologies, and engage stakeholders and industry, in an integrated effort to validate and implement technologies for better detection of foodborne hazards along the food production continuum.

Progress Report
The Center for Food Safety Engineering continues to develop novel methods for the sampling and detection of pathogens and other organisms that are indicators of contamination with pathogens, new methods to characterize microbial communities associated with food, and develop sensor technologies and bacterial growth models that will allow environmental data to be used to predict food safety risk. Despite the impact of COVID-19, most of our projects were able to achieve their 60-month milestones. Work has continued on development of elastic light scattering (ELS) for the identification of both pathogenic and non-pathogenic indicator organisms. The hyperspectral version of ELS (HESPI) is undergoing testing with bacterial species that have different colony morphologies, with predictions from theoretical models being compared to scatter patterns achieved with different wavelengths of light. Work on using ELS to classify entire microbial communities (bacterial and fungal) is leading to improvements in software to allow the system to classify colonies as unknown (rather than having to be classified as something in the current library) and also to define the confidence with which an assignment is made. Our microbial community analyses have utilized next-gen sequencing to identify bacteria and fungi found within the endogenous community on romaine lettuce. Community analysis has been performed on conventionally and organically grown commercial lettuce. This work has identified a yeast species that appears to be ubiquitously found on lettuce and that we are working to modify into a living biosensor to detect human pathogens. We also conducted experiments to identify factors that promote or restrict the growth of E. coli O157:H7 when introduced to lettuce leaves. These experiments demonstrate that environmental factors (particularly relative humidity) play a major role in the ability of the pathogen to persist in the lettuce phylloplane community. These same environmental factors also affect the composition and density of the bacterial community and allow the identification of indicator species that are correlated with higher and lower growth of E. coli O157:H7. We continue to develop Laser Induced Breakdown Spectroscopy (LIBS) hardware and software to detect biological and chemical contaminants in food. This work has focused both on a lateral flow assay (LFA)-based system using lanthanide labeled antibodies and a handheld system that will produce “food-fingerprints” that can be used to detect adulteration and food fraud. The second system will take advantage of our work on machine-learning models, with a specific focus on the interpretation and classification of spectral data linking LIBS food fingerprinting with elemental analysis. We have expanded our machine learning (ML) models to include data from olive oil, balsamic vinegar, and coffee, in addition to alpine-style cheeses. Two routes of explainable ML have been pursued: 1) we continued using generalized linear models with various regularization approaches to select spatial features, and 2) we build a feature dependency map using a graph-theoretic approach paired with manifold learning to identify the spectral features that contribute most to the fingerprint distinguishability. We have developed a “biphasic” amplification method for E. coli detection in ground beef samples and demonstrated detection of E. coli O157:H7 DNA with a limit of 2.5 copies in 30 mg of meat sample (~83.3 copies/gram) as well as E. coli O157:H7 cells with a detection limit of 1 CFU/30 mg of meat sample (~33 CFU/g). The platform is based on a two-phase reaction in which the solid phase is dried food matrix that does not remix with the supernatant phase where the fluorescent amplicons are concentrated, allowing a high signal to noise ratio and greater fluorescence change. The advantage of drying the food matrix is that amplification reagents such as primers and polymerase can detect the target without inhibition from reaction inhibitory factors that are locked in the dried matrix. This system provides a culture independent method where the current sample to result time is 2.5 hours and minimal sample processing assures the integrity of DNA or bacteria. We have developed patterned microfluidic paper-based analytical devices (µ-PADS) that have combined the well-known advantages of paper strips with the functionality and utility of microfluidics. This technology holds great potential for instrument-free, portable and multiplexed detection. These devices take advantage of gold nanoparticle’s (AuNPs) decorated on polystyrene microparticles (Au-PS). Highly selective aptamers were immobilized on AuNPs and act as a protective support material that maintains the Au-PS particles in a non-aggregated state, giving off a pink color visible to the naked eye. In the presence of the analyte the aptamer coating is disrupted leading to a stronger aptamer-target bond compared to the aptamer-particle bond. Upon target binding the aptamer is lifted off of the particles’ surface, causing their immediate agglomeration, which triggers a change in color from pink to purple. After optimizing the synthesis and agglomeration conditions of the polystyrene-AuNPs-aptamer complex, we were able to obtain change in color that can be seen with the naked eye in concentrations as low as 1 ppm. Preliminary detection results performed in solution showed sensitivity as low as 0.5 ppm for mercury and selectivity against other heavy metals such as arsenic, cadmium, and lead. In order to move forward with even more tests at the same time, we would like to create new devices with six to eight test zones. Our fabrication method for the eight-channel devices is wax printing due to the fact that we wish to keep part costs low and fabrication easy. Calibration of the wax-based printing methods is now complete, and multi-target tests are underway. To facilitate quantitative measurements from µ-PAD devices we have designed a pipeline for image analysis using a mobile phone camera to estimate target concentration. In this approach, the metric that we use to characterize the responses in each circular detection zone is based on the average grayscale value. This base model shows an average prediction performance of 60%. Its effectiveness is restricted by the limited dataset and the insufficient utilization of the spatial information contained in the sensor pad images. The detection of lower contamination levels remains challenging due to the small number of data samples and large intra-class variance. To overcome this challenge and improve the prediction accuracy, we propose two more methods for image processing: (1) a Deep Learning Approach for Classifying Contamination Levels with Limited Samples, (2) a Spectral Imaging for Contaminant Levels Estimation. Our systems yield much higher classification accuracy (88%, and 87% average accuracy, respectively) than the base model, demonstrating the feasibility of our approaches. To facilitate development of our phage-based detection systems we characterized genes previously isolated in the project based on their expression in inducible genetic systems. The expression of the phage repressor protein reduced the induction of the lytic cycle in E. coli O157:H7, reducing phage titers by greater than 99.9% and totally inhibiting plaque formation. Expression of the antirepressor showed the opposite effect by limiting the induction of the lysogenic cycle resulting in a primary lytic phenotype observed in plaques/titers even in our modified phage engineered to have a high frequency of lysogen formation. We have continued our work with the HEK293 reporter cell line expressing TLR-5 receptors. Interaction of flagellar antigen with TLR-5 leads to the activation of NF-kappaB, IL-8 and alkaline phosphatase which can be detected with an appropriate substrate to generate a blue color. The assay is highly specific and validated with the top 20 Salmonella enterica serovars and 10 non-Salmonella spp. The performance of the assay was also validated with spiked food samples. The total detection time including shortened pre-enrichment (4 h) and selective enrichment (4 h) set up with artificially inoculated outbreak-implicated food samples (chicken thighs, peanut kernel, peanut butter, black pepper, mayonnaise and peach) was 15 h when inoculated at 1–100 CFU/25 g sample. We have performed several analytical studies for the development of the battery-less time-temperature monitoring (TTM) system. This configuration will greatly simplify integration of the system into deli cases. It also does not depend on the power status of the deli cases, providing reports even during unscheduled power outages. The proposed scheme relies on energy harvesting of radiofrequency waves that are abundant in the environment or provided by internal oscillators that operate in the required frequency range. To estimate the required harvested power, a study of the power budget was performed for every essential module of our system. We also performed sensor resolution analysis for each excitation scheme finding that with internal oscillation the temperature resolution increases by 40% to a nominal value of 0.6 °C. Our next steps are to construct the new design and validate our results in the lab as well as in the field. Growth studies required the coupling of new sensors for routine monitoring of product temperature abuse over time and, unfortunately, sensor availability and testing was delayed due to Covid-19. Sensor performance was assessed by incorporating into in-lab incubators in parallel with built-in temperature monitoring. Growth curves of L. monocytogenes 10403S were completed in different brands of prepacked deli-style turkey breast containing celery powder as an antimicrobial. These data are being used to model the effect of temperature abuse on growth of L. monocytogenes and the utility of celery powder as a bacterial growth inhibitor.

1. Long term storage of reporter bacteriophages. Bacteriophages (viruses that only infect bacteria) offer a unique opportunity to expand detection technologies targeting foodborne pathogens. These bacteriophage-based technologies can potentially be integrated into standardized food testing protocols with minimal disruption of workflow. While several research groups have expanded the uses for various bacteriophages, reliable means of translating these proof-of-principle systems to impactful are lacking. The research accomplished by ARS-funded scientists at the Center for Food Safety Engineering in West Lafayette, Indiana, has developed technologies that allow bacteriophages for pathogen detection to be manufactured and stored at room temperature. Using bacteriophages reduces the time it takes to evaluate the safety of various food products, increasing the throughput of testing, reducing financial losses to food companies, and, most importantly, protecting consumers.

2. Cellphone-based portable fluorometer. A major limitation of handheld instruments for optical detection of foodborne pathogens is reliably and reproducibly of very weak signals. ARS-funded scientists at the Center for Food Safety Engineering in West Lafayette, Indiana, have developed a smartphone-based fluorescence detection system that can detect an extremely low fluorescence signal level. The system combines an optimized hardware design and software algorithm that enhances the low-level fluorescence signal to overcome these limitations. The incorporation of these new technologies will potentially enable detection systems that inspectors can use to provide on-site detection results.

3. Culture-free approach detection of pathogens in complex food samples. Rapid detection of the pathogen E. coli O157:H7 is vital for public health and food safety worldwide. It is very important to detect pathogen presence in food products early, rapidly, and accurately to avoid potential outbreaks and economic loss. ARS-funded scientists at the Center for Food Safety Engineering at the University of Illinois, Urbana-Champaign, Illinois, have developed a diagnostic tool to detect pathogens from direct samples without the necessary steps of enrichment or purification for the pathogens. The protocol directly detects pathogen DNA from a dried food sample and reduces processing time which improves the overall time to results to only 2.5 hours, whereas standard protocols of plating take 1-5 days. Experimental results showed that the method is very sensitive (a detection limit of 1 cfu/ 30mg of dried food matrix for E. coli bacteria). This technique bypasses the need for conventional steps of culture or nucleic acid purification, and the platform can reduce the costs and resources required for food testing and safety.

4. Highly sensitive rapid detection technology for monitoring pathogens in large volume food samples. ARS-funded scientists at the Center for Food Safety Engineering, the University of Illinois at Urbana-Champaign, Urbana, Illinois, have developed sensor technologies utilizing polyester pads to detect E. coli O157:H7 in large sample volumes of a complex food matrix by monitoring the change in color of a sensor strip that the naked eye can see. The method requires only a simple syringe-based setup with an antibody-modified polyester pad to make the enrichment approach practical. The developed method could process a large volume of sample (50 ml of lettuce cocktail blended from a 5g of lettuce), required only 40 minutes of sample processing time, and was very sensitive and specific (100 CFU/ml of E. coli O157:H7 in the presence of ~3000 CFU/ml of non-target bacteria). The developed approach demonstrates a promising means to improve the detection of target bacteria with a high degree of sensitivity and specificity. It could be used in low-resource settings for the specific detection of pathogens.

5. Celery powder to control Listeria monocytogenes in prepacked deli-style turkey breast. Listeria monocytogenes can be a post-processing contaminant of ready-to-eat foods, including deli meats. Deli meats are among the highest risk foods resulting in listeriosis as they do not undergo additional cooking steps before consumption. ARS-funded scientists at the Center for Food Safety Engineering in West Lafayette, Indiana, conducted growth studies to evaluate the efficacy of commercial celery powder as an antimicrobial against L. monocytogenes in prepacked deli-style turkey breast stored under ideal and temperature abuse conditions. The results demonstrated a significant linear relationship between product temperature abuse over time and the growth of L. monocytogenes in deli-style turkey samples. The findings underscore the importance of maintaining refrigeration temperatures to complement the efficacy of commercial antimicrobials.

6. Better methods for sequencing the bacterial microbiome from plant samples. Characterization of the bacterial communities from plant tissue is a valuable tool in produce safety research. Methods to catalog these bacterial communities require high throughput DNA sequencing of a gene present in all bacteria, the 16S rRNA gene. Characterization of the bacterial communities from plant tissue using high throughput DNA sequencing of the 16S rDNA is a challenge due to the high similarity between the plant mitochondrial and plastid DNA and the bacterial 16S rDNA. This can result in a greatly reduced number of bacterial sequence reads, leading to loss of the entire bacterial diversity coverage. ARS-funded scientists at the Center for Food Safety Engineering in West Lafayette, Indiana, have developed protocols for sample preparation, high throughput DNA sequencing, and data analysis to overcome this issue. These protocols will result in much more complete descriptions of plant bacterial communities at a lower cost than was previously possible.

7. Identification of environmental factors and biomarkers associated with E. coli O157:H7 growth on romaine lettuce leaves. Indoor farming industries grow leafy greens under controlled environmental conditions, potentially allowing them to restrict the growth of harmful bacteria on the leaf surface. ARS-funded scientists at the Center for Food Safety Engineering in West Lafayette, Indiana, have found that relative humidity (R.H.) is the main factor modulating the fate of E. coli O157:H7 and the composition of the resident bacterial microbiome on romaine lettuce. Other minor contributing factors included the inoculum dose of the foodborne pathogen and the wettability of the leaf surface. Two bacterial species were also found to be potential biomarkers for higher and lower growth rates of E. coli O157:H7 on lettuce leaves. Application of these results could increase the safety of indoor lettuce growing systems.

8. Deep learning approach for classifying contamination levels with limited samples. Mercury (Hg) and Arsenic (As) ions have been recognized as chemical threats to human health and can be present in food samples in trace amounts. Still, the detection of low levels of contamination remains challenging due to the small number of available data samples and significant intra-class variance. To overcome this challenge, ARS-funded scientists at the Center for Food Safety Engineering in West Lafayette, Indiana, explored techniques for synthesizing realistic colorimetric images and propose a Convolutional Neural Network (CNN) classifier for five-contamination-levels. The system was trained and evaluated on a limited dataset of 126 images captured with a cell phone camera representing five contamination levels. The system yields 88.1% classification accuracy and 91.9% precision, demonstrating the feasibility of this approach. The application of this system would allow the use of readily available cell phone cameras to capture images that can estimate the level of contamination with heavy metals.

Review Publications
Kanach, A., Bottorff, T., Zhao, M., Wang, J., Chiu, G.T., Applegate, B. 2021. Evaluation of anhydrous processing and storage methods of the temperate bacteriophage V10 for integration into foodborne pathogen detection methodologies. PLoS ONE. 16(4):e0249473.
Wang, J., Kanach, A., Han, R., Applegate, B. 2021. Application of bacteriophage in rapid detection of Escherichia coli in foods. Current Opinion in Food Science. 39:43–50.