SPREAD OF ESCHERICHIA COLI O157:H7 DURING FLUME WASHING AND DRYING OF FRESH-CUT ROMAINE LETTUCE. Siyi Wang

SPREAD OF ESCHERICHIA COLI O157:H7 DURING FLUME WASHING AND DRYING OF FRESH-CUT ROMAINE LETTUCE By Siyi Wang A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Food Science Master of Science 2016

ABSTRACT SPREAD OF ESCHERICHIA COLI O157:H7 DURING FLUME WASHING AND DRYING OF FRESH-CUT ROMAINE LETTUCE By Siyi Wang The microbiological safety of leafy greens remains a concern as evidenced from recent outbreaks. This study assessed the spread of E. coli O157:H7 during washing and drying of fresh-cut romaine lettuce. Radicchio was spot-inoculated at 10-1, 10 1 and 10 3 CFU/leaf and mixed with uninoculated romaine lettuce to obtain 5 kg batches with inoculated vs uninoculated ratios of 0.5:100, 1:100, 5:100 and 10:100. After 90 s of sanitizer-free flume washing followed by shaker table and centrifugal dryer, the radicchio was removed and the lettuce was divided into 225 g samples to test presence/absence of E. coli O157:H7 using a GeneQuence assay. Based on triplicate trials, lower inoculation levels led to decreased E. coli O157:H7 transfer to romaine lettuce (P < 0.05). All lettuce samples yielded E. coli O157:H7 when radicchio was inoculated at 10 3 CFU/leaf. At 10 1 CFU/leaf, the percentage of positive samples decreased from 96.8% to 93.7%, 81.0% and 63.5% while at 10-1 CFU/leaf, 22.2%, 6.3%, 4.8% and 6.3% were positive at 10:100, 5:100, 1:100 and 0.5:100 ratios. Within each inoculation level, there were no significant differences (P>0.05) among four product ratios. These findings are critical to predict the extent of cross-contamination under realistic conditions and will provide important data for improving exposure assessment in risk assessments for leafy greens.

To my mom Yan Cheng, my dad Xiaodong Wang iii

ACKNOWLEDGEMENTS Here I would like to say thank you to my major advisor, Dr. Elliot Ryser, who brought me into the food safety field. He offered me the opportunity to volunteer in the lab as an undergraduate student assistant, which opened the door to my commitment to scientific study and learning. It is my great pleasure to learn with and work for him. Without his support and guidelines, I would not have achieved all these accomplishments by myself. I also would like to thank my committee members Dr. Cornelius Barry and Dr. John Linz for their advice and support keeping me on the right track for my research. I want to thank Haley Smolinski and Lin Ren for all the assistance and being great lab partners and friends. Without their help, it would have taken much longer to finish the project. As well, I would like to especially thank Hamoud Alnughaymishi, Ryann Gustafson and all my lab mates for their help, sharing and friendship. Last, but not least, I would like to thank my parents for their love. Siyi Wang iv

TABLE OF CONTENTS LIST OF TABLES... vii LIST OF FIGURES... viii KEY TO SYMBOLS AND ABBREVIATIONS... x INTRODUCTION... 1 CHAPTER 1: Review of Pertinent Literature... 4 1.1 Fresh-cut production and lettuce consumption.... 5 1.2 Outbreaks and safety concerns associated with the fresh-cut industry..... 8 1.3 Escherichia coli O157:H7....... 11 1.4 Sources of contamination during pre-harvest and post-harvest handling of lettuce 12 1.4.1 Pre-harvest contamination... 13 1.4.2 Post-harvest contamination...... 14 1.5 Post-harvest processing and sources of contamination.... 16 1.5.1 Shredding......17 1.5.2 Conveying.....18 1.5.3 Flume Washing.... 18 1.5.4 Dewatering and Drying.... 20 1.6 Previous microbial studies on fresh produce with spot inoculation.. 20 1.7 Previous cross-contamination studies on pathogen transfer and redistribution during processing..... 22 1.8 Risk Analysis & Assessment......... 23 1.9 FSMA and the influences on food safety.... 24 1.10 Overall goals..... 25 CHAPTER 2: Spread of Escherichia coli O157:H7 during Flume Washing and Drying of Fresh Cut Romaine Lettuce... 26 2.1 OBJECTIVE... 27 2.2 MATERIALS AND METHODS... 28 2.2.1 Overall experimental design... 28 2.2.2 Produce... 29 2.2.3 Bacterial strains used... 29 2.2.4 Inoculation of radicchio... 30 v

2.2.5 Processing line... 31 2.2.6 Romaine lettuce processing and sample collection... 32 2.2.7 Chemical sanitizers... 35 2.2.8 Microbiological analysis... 35 2.2.9 GeneQuence Assay... 36 2.2.10 Statistical analysis.... 38 2.3 RESULTS... 39 2.3.1 Romaine lettuce....... 39 2.3.2 Radicchio........ 42 2.3.3 Water...... 45 2.3.4 Sanitizer trials.......... 45 2.4 DISCUSSION... 47 CHAPTER 3: Conclusions and Future Recommendations... 53 APPENDIX... 57 BIBLIOGRAPHY... 67 vi

LIST OF TABLES Table 1.1: Fresh produce including leafy green-associated E. coli O157:H7 outbreaks since 2006 based on Foodborne Outbreak Online Database......... 9 Table 2.1: Numbers and percent of positive samples among ~20 total samples with different inoculation levels and product ratios..... 41 Table 2.2: E. coli O157:H7 populations on radicchio before washing and E. coli O157:H7 log reductions on radicchio after processing. 44 Table 2.3: Comparison between sanitizer-free trials and 60 ppm sanitizer trials at the same inoculation level (10 3 CFU/leaf) and inoculated:uninoculated ratio (10:100)..... 46 Table AI.1: Percent of samples detected as positive for E. coli O157:H7 due to the E. coli O157:H7 transfer to 225 g samples of romaine lettuce after 90 sec of flume washing in sanitizer-free water... 59 Table AI.2: E. coli O157:H7 populations on radicchio leaves before processing (log CFU/g)... 60 Table AI.3: E. coli O157:H7 population on radicchio leaves after processing (log CFU/g). 61 Table AI.4: Table AI.4: Percent of E. coli O157:H7 shed from radicchio leaves after processing. 62 Table AI.5: E. coli O157:H7 populations in wash water sample taken at the end of processing (log CFU/100 ml).... 63 Table AI.6: E. coli O157:H7 populations in water sample for each centrifugal drying batch taken (log CFU/100 ml)..64 Table AI.7: Summary of the results for experiments using 60 ppm sanitizer..... 65 vii

LIST OF FIGURES Figure 1.1: Percentage of Vegetable and Legume Availability in 2013.... 7 Figure 1.2: Lettuce U.S. imports.. 7 Figure 2.1: Lettuce shredding..... 33 Figure 2.2: Flume washing and dewatering.... 33 Figure 2.3: Centrifugation... 34 Figure 2.4: Sample collection..... 34 Figure 2.5: The Stat Fax 4200 microplate reader and Stat Fax 2600 semi-automatic microplate washer... 37 Figure 2.6: GeneQuence kit 37 Figure 2.7: Percent of samples detected as positive for E. coli O157:H7 due to the E. coli O157:H7 transfer to 225 g samples of romaine lettuce after 90 sec of flume washing in sanitizer-free water based on different inoculation levels. Mean values with different letters are significantly different (P 0.05) within the same product ratio (10:100, 5:100, 1:100, 0.5:100).......... 40 Figure 2.8: Percent of samples detected as positive for E. coli O157:H7 due to the E. coli O157:H7 transfer to 225g samples of romaine lettuce after 90 sec of flume washing in sanitizer-free water based on different product inoculated:uninoculated ratios. Mean values with same letter are not significantly different (P > 0.05) within the same inoculation level.. 41 Figure 2.9: E. coli O157:H7 population on radicchio leaves before and after processing based on different inoculated:uninoculated product ratios. The log reductions after washing are not significantly different within the same inoculation level (P > 0.05) 43 viii

Figure 2.10: E. coli O157:H7 population on radicchio leaves before and after processing based on different inoculation level. At 10-1 CFU/leaf, E. coli populations are below limit of detection (1 CFU/25g) so percent loss is not applicable at this inoculation level. The log reductions after washing are not significantly different within the same produce ratio (P > 0.05). 43 Figure AI.1: Growth curves in GeneQuence enrichment during 24 h incubation with different product:enrichment ratio (1:2 vs 1:9)... 66 ix

KEY TO SYMBOLS AND ABBREVIATIONS ANOVA Analysis of Variance AgMRC Agriculture Marketing Resource Center CDC Centers for Disease Control and Prevention CFU Colony forming unit(s) CFSAN Center for Food Safety and Applied Nutrition LGMA California Leafy Green Products Handler Marketing Agreement cm centimeter(s) CSPI Center for Science in the Public Interest ERS Economic Research Service EHEC Enterohemorrhagic Escherichia coli FDA Food and Drug Administration g gram(s) FSMA Food Safety Modernization Act GAPs Good Agricultural Practices GFP-labeled Green Fluorescent Protein-labeled GMPs Good Manufacturing Practices HUS Hemolytic uremic syndrome kg kilometer(s) L liter(s) LOD limit of detection ml milliliter(s) x

PBS Phosphate Buffered Saline ppm Parts per million QPRAM Quantitative Predictive Risk Assessment Model RTE Ready-to-eat RH Relative Humidity s second(s) SD Standard Deviation ELISA Sandwich Enzyme-linked Immunosorbent Assay SSOPs Sanitation Standard Operation Procedures USDA United States Department of Agriculture TSA-YE Trypticase Soy Agar with 0.6% Yeast Extract TSB-YE Trypticase Soy Broth with 0.6% Yeast Extract US United States of America µl microliter(s) xi

INTRODUCTION 1

Nowadays, fresh produce consumption in the United States has increased significantly. As an important source of vitamins, nutrients and fiber, the high demand for fresh produce can be explained by the increased consumer interest in developing healthy eating habits. Due to advances in preservation technology and benefits of global trade, many varieties of fresh produce are available all year long (Olaimat and Holley, 2012). Edible coatings and films with added antimicrobial compounds are also being used more frequently in the food industry to preserve fresh fruits and vegetables and increase shelf life (Silvia et al., 2011). Even though consumers benefit from the rapid growth of fresh produce consumption by having a healthier lifestyle, at the same time, no one can ignore the fact that there are more foodborne illness outbreaks associated with fresh produce consumption, especially leafy greens (Warriner et al., 2009). In recent years, the contamination of leafy greens associated with E. coli O157: H7 has drawn significant attention from government and industry, as well as researchers (Danyluk and Schaffner, 2011). Leafy greens consumed raw, such as lettuce, are most likely to be connected with diseases resulting from E. coli O157:H7 contamination (Franz et al., 2009). According to the Centers for Disease Control and Prevention (CDC), multiple outbreaks of E. coli O157:H7 infections have been linked to leafy greens, which causes a huge economic loss and human illness. For example, one multistate outbreak of E. coli O157:H7 infection associated with ready-to-eat salads resulted in a total of 33 infections including 13 hospitalizations in four states, with 7 hospitalizations and 0 death (CDC, 2016). To prevent these foodborne disease outbreaks, it is necessary to understand the mechanism of cross-contamination in the food chain. 2

Cross-contamination may occur at any point from farm to fork. The sources of contamination include contaminated irrigation water and soil, as well as improper human handling (Beuchat, 2002). When one leaf of lettuce is contaminated with E. coli O157:H7 in the field, this pathogen can transfer to other leaves, the wash water or equipment during commercial processing in the plant, leading to large quantities of cross-contaminated product. As food supply chains are getting more complex and global, the Food and Drug Administration (FDA) uses risk analysis to minimize the risk of contamination and illness (FDA, 2016). To ensure food safety, risk assessments are being used to better understand the interactions between hazards, foods and human hosts. Besides pre-harvest cross-contamination, it is also important to understand how contamination can be spread in the processing pathway. Since the extent to which cross-contamination occurs is not well characterized, additional research to quantify pathogen redistribution during post-harvest processing is needed in order to fill the current data gaps. Consequently, the objectives of this study were to: 1. Quantify the spread of Escherichia coli O157:H7 during pilot-scale sanitizer-free washing and drying of fresh-cut romaine lettuce. 2. Assess the spread of Escherichia coli O157:H7 during processing with chemical sanitizers. 3

CHAPTER 1: Review of Pertinent Literature 4

1.1 Fresh-cut production and lettuce consumption In recent years, consumption of fresh produce has increased significantly due to healthy eating trends. As a very important portion of the Americans diet, fresh produce provides a high level of vitamins and minerals that contribute to human health. Consequently, the demand for fresh-cut leafy green vegetables such as lettuce has continued to expand because more consumers want to achieve healthy diets and convenience at the same time (Weng et al., 2016). According to the United States Department of Agriculture (USDA) (2016), fresh lettuce, including head, romaine and leaf lettuce, was one of the top three vegetables consumed in the US. Based on the Percentage of Vegetable and Legume Availability Chart in 2013 (Figure 1.1), the consumption was 25.5 pounds per person in 2013. With $12 billion in annual sales in the past few years in the US, the fresh-cut sector of the produce industry has become the fastest growing segment (FDA, 2008). In general, there are two types of lettuce: head lettuce and leaf lettuce, which includes romaine, butter-head and leaf types. Even though lettuce is produced in all 50 states, the top two states, California and Arizona, accounted for 98 percent of the leaf lettuce in 2013 nationwide. Other than local production, US imports of lettuce have increased in recent years according to the Vegetables and Pulses Data in Figure 1.2 (ERS, 2016). As fresh products, almost all head lettuce is field-packed for bulk sale or transported to salad processing facilities (Agriculture Marketing Resource Center (AgMRC), 2016). To be minimally processed, lettuce must be washed, chopped into small pieces, mixed with other leafy greens, sorted into bags and sold as 5

salad products in the market. Compared to raw vegetables, minimally processed produce is more perishable with a relatively short shelf life (Siroli et al., 2015). 6

Figure 1.1: Percentage of Vegetable and Legume Availability in 2013 (USDA, 2013) Figure 1.2: Lettuce U.S. imports (ERS, 2016) 7

1.2 Outbreaks and safety concerns associated with the fresh-cut industry Normally, since there is no inactivation step before consuming fresh-cut products such as ready-to-eat (RTE) salads, this type of product is likely to become a potential source of human pathogens (Castro-Ibanez et al., 2015). As the market continues to grow, various challenges have emerged along the production chain from the raw material to its consumption to ensure food quality and safety. Due to changes in food production and consumption patterns, produce-associated outbreaks have been reported more frequently. Because the fresh-cut industry moved from small local farms to large-scaled central processing facilities, the negative impact on the industry can be broader (Buchholz et al., 2014). From 1996 to 2006, there were 72 foodborne illness outbreaks associated with the consumption of fresh-cut produce. It is also noted that lettuce and other leafy greens have been implicated in multiple foodborne disease outbreaks especially associated with E. coli O157:H7 (Jensen et al., 2014). The outbreaks of fresh produce, including leafy greens associated with E. coli O157:H7 since 2006, are summarized in Table 1.1 (CDC, 2016). In 2015, there was a multistate outbreak of listeriosis associated with packaged salads produced in an Ohio processing plant, which caused 19 hospitalizations and one death (CDC, 2016). According to the Center for Science in the Public Interest (CSPI), the number of outbreaks linked to fresh produce consumption, especially lettuce, has increased since 2000 (CSPI, 2009). Those outbreaks not only affect public health, but also lead to economic losses for the society at the same time. 8

Table 1.1: Fresh produce including leafy green-associated E. coli O157:H7 outbreaks since 2006 based on Foodborne Outbreak Online Database (CDC, 2016). Year Location Illnesses reported Produce 2006 Multistate 238(3 deaths) Spinach 2006 Multistate 77 Lettuce 2006 Multistate 80 Lettuce 2007 AL 26 Salads 2008 WA 10 Spinach 2008 Multistate 13 Spinach 2009 Multistate 16 Lettuce 2009 Multistate 22 Lettuce 2010 Multistate 31 Romaine Lettuce 2011 Multistate 26 Lettuce 2011 Multistate 60 Romaine Lettuce 2012 CA 12 Lettuce 2012 Multistate 58 (3 developed HUS) Romaine Lettuce 9

Table 1.1 (cont d) 2012 Multistate 33 Packaged leafy greens 2012 Multistate 16 Lettuce 2013 Multistate 33 Salads 2016 Multistate 9 Alfalfa Sprouts Even though eating vegetables raw helps maintain the nutrients in fresh produce, this practice may also can be risky due to the presence of pathogens. Vegetables can become contaminated during cultivation, harvest or postharvest with foodborne pathogens. Two major pathogens associated with vegetables are Salmonella and Shiga toxin-producing Escherichia coli, which can be introduced through soil, irrigation water, insects and human handling (Delbeke et al., 2014). With improper handling or storage conditions, such contamination can lead to serious consequences. According to the FDA, processing whole fresh produce into fresh-cut products increases the risk of bacterial growth and contamination. This is because the natural exterior barrier of the produce is broken during processing. When produce is chopped or shredded, the release of plant cellular fluids provides nutrients for bacterial growth. In addition, the cross-contamination that occurs during fresh-cut processing may put large volumes of product at risk (FDA, 2008). 10

1.3 Escherichia coli O157:H7 Escherichia coli is a Gram-negative, rod-shaped, facultative anaerobic bacterium. While most E. coli strains colonize the GI tract of humans and animals harmlessly, there are some pathogenic strains which acquire virulence factors through plasmids, transposons, bacteriophages, and/or pathogenicity islands (Lim et al., 2010). As a subgroup of E. coli, enterohemorrhagic E. coli (EHEC) may cause bloody diarrhea, hemorrhagic colitis and hemolytic uremic syndrome (HUS) (Goswami et al., 2015). Among EHEC strains, E. coli O157:H7 is a leading cause of foodborne and waterborne disease outbreaks in the US with other outbreaks associated with E. coli O157:H7 having been reported worldwide including Australia, Canada, Japan and various countries in Europe and southern Africa (Huang et al., 2014). As one of the major virulence factors involved in E. coli O157:H7 pathogenesis, the Shiga toxins (Stx) are classified as either Stx1 or Stx 2 based on their immunoreactivity (Yin et al., 2015). Other than E. coli O157:H7, many other serotypes of STEC cause disease including E. coli O145. Limited public health surveillance data exists for these other serotypes with many such infections going unreported or undiagnosed (CDC, 2015). E. coli O157:H7 was first recognized in 1982 as a human pathogen associated with outbreaks of bloody diarrhea in Oregon and Michigan (Lim et al., 2010). The first outbreak occurred in Oregon with 26 cases of which 19 were hospitalized. After three months, the second outbreak was reported in Michigan with 21 cases and 14 hospitalizations (Institute of Food Technologists, 1997). In 2006, a multistate outbreak associated with E. coli O157:H7 linked to 11

spinach consumption was identified, causing over 200 cases and three deaths (Gelting et al., 2011). Based on information published by the FDA, the infective dose is still unknown but may be similar to that of Shigella spp. with as few as 10 organisms inducing illness (FDA, 2014). Human illness may be observed even after ingestion of 1 CFU and disease in humans may develop without prior multiplication in food (Boqvist et al., 2015). Symptoms observed from most identified cases include severe diarrhea and abdominal cramps. Even though the target populations include people of all ages, young children and the elderly seem to be most susceptible to developing more serious symptoms. Particularly for children under five years old, the infection may cause a complication called hemolytic uremic syndrome, which is a serious disease in which red blood cells are destroyed and the kidneys fail due to platelet aggregation in renal arteries and glomerular capillaries (NY State Department of Health, 2006). 1.4 Sources of contamination during pre-harvest and post-harvest handling of lettuce Even though spoilage bacteria, yeasts and molds predominate on fresh produce such as leafy greens, fresh produce can still be contaminated with human pathogens, parasites and viruses from the stages of growing, harvesting, postharvest handling, processing and distribution (Gil et al., 2015; Olaimat and Holley, 2012). As we know, there are various sources of contamination that occur at pre-harvest or post-harvest steps before and after transferring to processing facilities. 12

1.4.1 Pre-harvest contamination From 1996 to 2008, almost half of the fresh produce outbreaks were traced to leafy greens containing E. coli O157:H7 with implications of pre-harvest contamination (D Lima and Suslow, 2009). At the pre-harvest stage of lettuce production, sources of contamination may include soil, feces, irrigation water, insecticides, dust, insects, inadequately composted manure and wild or domestic animals (Beuchat, 2002). As well, improper human handling such as using unclean tools in the field can also introduce contamination. For instance, when lettuce is cut and cored by workers in the field, blades of field coring devices may contact contaminated soil and transfer pathogens to lettuce tissues, which causes cross-contamination. According to a simulated field coring study, when using the same contaminated field coring device to consecutively cut the stems of the lettuce heads, E. coli O157:H7 can be transferred from one head to the next (Taormina et al., 2008). Soil can be considered as a natural environment for human pathogens especially when animal wastes are added for fertilizers. Based on multiple reports, E. coli O157:H7 is able to survive in soil for 7 to 25 weeks, depending on the soil type and conditions, which makes cross-contamination more likely to occur (Lang and Smith, 2007; Olaimat and Holley, 2012). Microbial safety of lettuce can be affected by the quality of irrigation water and type of irrigation system (Olaimat and Holley, 2012). As reported by Solomon et al. (2002), spray irrigation represents a higher level of risk for lettuce plants to be contaminated compared to surface irrigation. As a result, when pathogens transfer from soil or water, surface contamination 13

of the edible lettuce leaves occurs, which increases the chance of foodborne disease (Delaquis et al., 2007). In addition, airborne insect pests such as fruit flies that harbor E. coli on their bodies or guts can contaminate the surface of lettuce and serve as primary vectors of human pathogens. Insects may also affect the behavior of pathogens and damage the lettuce surface indirectly through their feeding activities (Erickson et al., 2010). The survivability of pathogens depends on environmental factors and weather conditions including predation, competition, water stress, temperature, UV radiation, ph, inorganic ammonia and organic nutrients (Rogers and Haines, 2005). In order to minimize contamination of fresh produce such as lettuce by foodborne pathogens, both contamination routes and environmental factors should be considered when analyzing causes of cross-contaminations (FDA, 2006; Park et al., 2014). 1.4.2 Post-harvest contamination After harvesting, lettuce is typically hand-cut, occasionally field cored, packed in the field and placed in large bulk bins for transport to the precooling or processing facilities (Delaquis et al., 2007). Any improper handling after harvest may lead to safety concerns. From previous studies, leafy greens such as lettuce can be contaminated during storage and transport (Gil et al., 2015). Consequently, lettuce should be stored in adequate facilities and transported in vehicles to minimize damage and access by pests (CAC/RCP 53, 2003). Any lettuce showing signs of decay 14

should be removed and discarded before transport and storage to minimize the risk of getting human pathogens associated with decay or damage (Brandl, 2008; FDA, 2006). During transport and storage, the preferred method used for precooling lettuce is vacuum cooling, considering its high surface-to-volume ratio. In vacuum cooling, produce is placed inside a vacuum chamber after which a vacuum is used to evaporate the water from the produce surface, lowering the temperature of the product (Ezeike and Hung, 2009). Optimum storage of lettuce is as close to 0 o C as possible with 98 to 100% relative humidity (RH), but the freezing injury may occur when stored at -0.2 o C (Saltveit, 2016; UC Davis, 2013). After transport from the field to the facility, lettuce should be stored as soon as possible. In general, if lettuce is stored at 5 o C or below, it will help prevent the growth of E. coli O157:H7. However, temperature abuse that occurs during storage and transportation may lead to problems (Khalil and Frank, 2010). Several studies have focused on the effect of temperature on the survival of E. coli O157: H7. Li et al. (2001) found a decline in E. coli O157:H7 populations on shredded iceberg lettuce at 5 o C, and an increase in growth during storage at 15 o C of 2.0 and 3.0 log CFU/g after 7 and 14 days, respectively. In another study conducted by Francis and O Brien (2001), the researchers reported that there was no change in growth of E. coli O157:H7 populations in cut iceberg lettuce when stored at 4 and 5 o C, while an increase of 1.0 log CFU/g was recorded at 8 and 10 o C after 12 days and 120 h, respectively. Hence, it is necessary to track temperature changes in order to minimize microbial growth during periods of temperature abuse. Additionally, good 15

hygiene, cleaning and sanitation practices during postharvest storage are necessary to ensure food safety (FDA, 2008). 1.5 Post-harvest processing and sources of contamination A significant amount of lettuce is processed into ready-to-eat, fresh-cut salads to fulfill consumers healthy eating desires. However, contamination of a small batch of lettuce can spread to large quantities of product during further processing, and may lead to foodborne disease outbreaks that affect the health of consumers. According to a recent outbreak investigation, the equipment used for cutting and shredding leafy greens in a processing plant was identified as one source of contamination (Stafford, 2002). Based on previous research conducted in a pilot plant-scale facility, one contaminated batch of leafy greens can easily contaminate subsequent uncontaminated batches if no effective antimicrobial interventions are applied (Buchholz et al., 2012). To prevent and minimize the chance of contamination, fresh produce processors should follow the FDA Food Safety Modernization Act (FSMA) Produce Safety rules, which provide standards for the growing, harvesting, packing, and holding of produce for human consumption (FDA, 2016), as well as the food safety guidelines for leafy greens as outlined in the California Leafy Green Products Handler Marketing Agreement (LGMA). In addition, sanitation programs such as Good Manufacturing Practices (GMPs), Good Agricultural Practices (GAPs), and Sanitation Standard Operation Procedures (SSOPs) should also be followed (FDA, 2006). Moreover, it is important 16

to understand how pathogens transfer during processing so preventive actions can be applied at each step to minimize contamination. After transport to processing facilities, the production of fresh-cut lettuce involves shredding, flume washing in water with sanitizers, shaker table dewatering, centrifugal drying to remove remaining water and packing into small bags (Buchholz et al., 2014). The processing equipment used can vary based on the size and product type of the facility, but for the most part it includes a shredder, conveyer, flume tank, shaker table and centrifugal dryer where cross-contamination may occur from any piece of equipment. 1.5.1 Shredding During the shredding step, lettuce is mechanically shredded using high-speed machines (Barry-Ryan and O Beirne, 1998). While shredding reduces the size of the product, the damage during cutting irrevocably alters biochemical characteristics of tissues and provides opportunities for microbial invasion (Delaquis et al., 2007). A study conducted by Khalil and Frank (2009) showed that the populations of E. coli O157:H7 on shredded lettuce pieces were significantly larger than the populations on intact leaves. Researchers in Japan found that fresh-cut vegetable samples from examined factories were heavily contaminated with E. coli O157:H7 from equipment surfaces used for trimming and slicing (Kaneko et al., 1999). Buchholz et al. (2012) assessed the transfer of E. coli O157:H7 from equipment surfaces to fresh-cut produce. During pilot plant- scale processing, E. coli O157:H7 populations decreased 0 to 3.6 log CFU/100 cm 2 on the equipment surface after 17

processing with the greatest losses seen for the shredder. These findings suggest a high likelihood of cross-contamination between shredders and produce. 1.5.2 Conveying Conveyors can be used at multiple points in the processing line to transport produce between different steps. After shredding, the shredded lettuce is step-conveyed and transported to the flume washer. There are various types of materials such as polypropylene and acetal that can be used for conveyor belts (Buchholz et al., 2010). While the surfaces of these conveyor belts tend to be smooth to maximize cleaning and sanitizing efficiency, there are still chances for bacteria to settle and proliferate. In a study conducted by Mafu et al. (1990) crevices and holes were observed on the surfaces of conveyor belts made of polypropylene, rubber and stainless steel, using scanning electron microscopy. Allen et al. (2005) showed that the Salmonella can persist up to 28 days on stainless steel, polyvinylchloride, and wooden surfaces during fall and winter (20 o C/60% RH) with undetectable levels on sponge rollers and conveyor belts after 7 and 21 days, respectively. Considering the long period of pathogen survival, proper sanitation programs should be followed closely to prevent contamination. 1.5.3 Flume Washing In general, fresh-cut lettuce is washed in cold water with sanitizers. This important step in processing removes dirt and exudates, improves product appearance, and reduces the microbial load, but can transfer pathogens to the water (Haute et al., 2015; Munther et al., 2015). Proper 18

decontamination procedures such as washing with sanitizers are critical to protect the safety of fresh-cut produce (Qadri et al., 2015). Among numerous types of sanitizers, chlorine is most widely used because of its low cost and ease of use. The antimicrobial activity of chlorine depends on the amount of free chlorine in the solution, the ph, the temperature, and the amount of organic matter (Suslow, 2001). Even though widely used, chlorine has been criticized for lacking efficacy against pathogens in certain conditions. Efficacy is limited in the presence of high organic loads, increased temperature, or light exposure (Gonzalez et al., 2004). For example, the chlorine compounds can be consumed with organic and inorganic constituents rapidly so that the efficacy of chlorine-based sanitizers is reduced (Zhou, 2013). Since more foodborne outbreaks have been associated with fresh-cut produce processing in recent years, many studies have focused on comparing and improving the antimicrobial activity of sanitizers. Lopez-Galvez et al. (2009) conducted an experiment to compare the efficacy of chlorine, Tsunami, Citrox and Purac against non-pathogenic E. coli on fresh-cut lettuce and the processing water used. The results showed that Citrox (5000 mg/l) and Purac (20000 mg/l) at the recommended doses did not prevent E. coli transfer between contaminated and non-contaminated produce while chlorine (40 mg/l) and Tsunami (500 mg/l) prevented cross-contamination during processing. Even though washing helps reduce the microbial load, the water used for washing can serve as a vehicle for cross-contamination. Pathogens can penetrate into crevices, cut surfaces or 19

intercellular spaces of lettuce, which become barriers for disinfection strategies (Buchholz et al., 2012). 1.5.4 Dewatering and Drying Following the washing step, fresh-cut lettuce is dewatered on a shaker table and dried using a centrifugal dryer to remove remaining water. This is an essential processing step because excess moisture may increase microbial growth (Cantwell and Suslow, 2004). In addition, this step must be performed with care to avoid tissue damage, product moisture content decrease and cell leakage that may lead to microbial growth on the produce (Artes and Allende, 2005). For different types of products, key parameters of the dewatering system such as time and speed of centrifugation must be adjusted properly. Especially for leafy greens, damage of tissues may occur due to high speed of centrifugation and cause products to leak fluids which reduces the quality (Artes and Allende, 2005; UC Davis, 2016). If the products are too delicate, then a forced air method should be applied (UC Davis, 2016). 1.6 Previous microbial studies on fresh produce with spot inoculation Based on many previous studies focused on microbial contamination of fresh produce, there are three common inoculation methods including spot inoculation, dip inoculation and spray inoculation. In spot inoculation, drops of a bacterial suspension are placed directly onto the produce surface using a micropipette (Koseki et al., 2003). Compared to the other inoculation methods, spot inoculation allows researchers to know the number of cells placed on the surface regardless of the weight or size of the product being tested 20

(Mukhopsdhyay et al., 2013). In dip inoculation, controlling the total volume of inoculum adhering to the produce surface remains difficult, which means the number of cells applied on the surface is unknown. In addition, spray inoculation also has issues regarding the entire amount of the spray contacting the surface, especially for smaller items (Lang et al., 2003). In one study, Lang et al. (2004) investigated how the survival and recovery of E. coli O157:H7 was affected by different methods of inoculation for lettuce and parsley. Compared to dip and spray inoculation, more consistent initial populations of pathogens were obtained after spot inoculating leafy greens, which allowed for more accurate measurement of microbial reductions. Another study that evaluated the efficacy of sanitizers concluded that the effectiveness of sanitizers against E. coli O157:H7 was affected by the inoculation method. It is possible that E. coli cells adhered less tenaciously to the surface with spot inoculation so treatments were more effective compared to spray or dip inoculation (Singh et al., 2002). Similarly, the experiment conducted on the evaluation of sanitizer efficacy of cantaloupes also showed that bacteria on spot-inoculated cantaloupes were more vulnerable due to the looser attachment to surfaces, which indicated that the bacteria can be detached more easily by washing treatments with sanitizers (Annous et al., 2005). Considering the advantages of spot inoculation discussed above and to better mimic field contamination of lettuce, spot inoculation was applied in this study. 21

1.7 Previous cross-contamination studies on pathogen transfer and redistribution during processing There are numerous studies indicating that cross-contamination can occur during fresh produce processing (Buchholz et al., 2012; Buchholz et al., 2014; Davidson et al., 2013; Wang and Ryser, 2013). Pathogens introduced by contaminated produce at earlier stages are likely to transfer and redistribute to pieces of equipment and the wash water upon further processing of, uncontaminated produce. It is important to track down how pathogens transfer during processing and better understand the extent of cross-contamination so that preventative actions can be developed and implemented. A quantitative study conducted by Holvoet et al. (2014) observed the cross-contamination processes and the transfer of E. coli from water to lettuce and vice versa without using sanitizers. This work showed the likelihood of cross-contamination at the flume washing step. Based on multiple experiments conducted in our laboratory, the cross-contamination pathway for leafy greens was determined through pilot plant trials. In one study by Buchholz et al. (2014), 45 kg of uninoculated lettuce was processed, followed by 9.1 kg of radicchio contaminated by dip inoculation with E. coli O157:H7 and finally by 907 kg of uninoculated iceberg lettuce, which was then collected into 40 bags (~22.7 kg per bag). Based on the results of direct plating with or without membrane filtration, E. coli was detected in lettuce, radicchio, and water samples and also found on equipment surface. Those findings demonstrate the potential for contaminated lettuce to spread pathogens to larger batches during processing. 22

In other work quantifying the transfer of E. coli O157:H7 to equipment during lettuce processing, 22.7 kg of iceberg lettuce was inoculated with different levels of E. coli O157:H7 and processed. The report showed that the produce contact surfaces of the shredder, conveyor, flume tank, shaker table and centrifuge were all contaminated to various degrees after processing (Buchholz et al., 2012). 1.8 Risk Analysis & Assessment According to the FDA, food safety decisions must be made based on available scientific data and systematic analysis to prevent contamination and illness (FDA, 2016). As food supplies become more global and complex, it is better to test large quantities of food to make sure there are no pathogens present. However, such large-scale testing is impossible in practice. Therefore, conducting quantitative microbial risk assessments using models and tools becomes a solution to enhance food safety (Rijgersberg et al., 2010). One quantitative risk assessment published for Listeria monocytogenes in ready-to-eat food predicted the relative risk rankings among 23 food categories based on two public health metrics outputs (Chen et al., 2013). As a risk analysis tool developed by the FDA, quantitative predictive risk assessment models (QPRAM) are used to predict and characterize risks from fresh produce consumption, and to track each unit of produce providing a history of details (FDA, 2015). These risk analysis tools provide researchers and federal agencies with an important strategy in preventing future contamination events. 23

The Codex Alimentarius points out that risk analysis is a framework used by food safety personnel to help with decision-making and to better understand the interactions between hazards, foods and human hosts (Demortain, 2012). Even though the FDA has developed a number of risk assessment tools such as the FDA-iRisk, due to the lack of available published data on cross-contamination studies of fresh-cut produce, large informational data gaps exist that limit the accuracy of the models being developed. Consequently, more data needs to be collected to improve the reliability of current risk assessments. 1.9 FSMA and the influences on food safety In 2011, the FDA Food Safety Modernization Act (FSMA), a set of regulations which expands the scope of food safety laws, was signed into law by President Obama, aimed at better protecting public health by improving food safety systems in the US. With these regulations, the focus shifted from simply responding to preventing contamination events. In 2016, the FDA made $19 million available to help states implement these food safety rules through activities such as educational programs, inspection and technical assistance (FDA, 2016; Pouliot, 2014). Specifically, FSMA encourages coordination and collaboration among federal, state and local agencies in the area of inspections and food safety (Tai, 2015). One section of the final rule focuses on produce safety. Quoting from the FDA, this rule establishes, for the first time, science-based minimum standards for the safe growing, harvesting, packing, and holding of fruits and vegetables grown for human consumption. In this case, the 24

risk of contamination is more likely to be reduced for fresh produce production and processing if these standards are followed. 1.9 Overall goals The microbiological safety of fresh-cut leafy greens remains an ongoing concern as evidenced by scores of recalls and sporadic outbreaks. Consequently, this study was designed to assess the transfer and redistribution of realistically low levels of Escherichia coli O157:H7 during simulated commercial production of fresh-cut romaine lettuce. Based on previous studies conducted in Dr. Ryser s laboratory, a small portion of contaminated product may spread and contaminate a large batch during processing (Buchholz et al., 2014). Consequently, a better understanding of how cross-contamination occurs in the processing pathway and the extent to which the cross-contamination occurs is needed. These findings, which reflect real-world contamination levels, will be critical to improving current risk assessment tools under development. 25

CHAPTER 2: Spread of Escherichia coli O157:H7 during Flume Washing and Drying of Fresh Cut Romaine Lettuce 26

2.1 OBJECTIVE The objective of this cross-contamination study was to quantify Escherichia coli O157:H7 transfer and redistribution on fresh-cut romaine lettuce and wash water during post-harvest processing, both with and without sanitizers. 27

2.2 MATERIALS AND METHODS 2.2.1 Overall experimental design. This study quantified the variability in redistribution of E. coli O157:H7 contamination after simulated commercial production of fresh-cut romaine lettuce. All transfer and redistribution data were obtained using Michigan State University s pilot-scale production line for fresh-cut leafy greens. The results were based on three replicates giving a total 36 experiments. Using inoculated radicchio as a colored surrogate for uninoculated romaine lettuce, radicchio leaves (~5 x 5 cm) were spot-inoculated one day ahead of processing and held overnight at 4 o C to achieve three inoculation levels (~10-1, 10 1 and 10 3 CFU/leaf). On the day of processing, the inoculated radicchio leaves were mixed with uninoculated romaine lettuce leaves (~5 x 5 cm) to obtain a 5 kg batch with inoculated:uninoculated product ratios of 0.5:100, 1:100, 5:100 and 10:100. After 90 seconds of flume washing in sanitizer-free water followed by shaker table dewatering and centrifugal drying, all radicchio leaves were removed from the lettuce before testing. Each batch of romaine lettuce (5kg) was then divided into ~20 225-g samples which were analyzed for presence/absence of E. coli O157:H7 using the E. coli O157:H7 GeneQuence asssay (Neogen Corp, Lansing, MI). The percentage of E. coli O157:H7 lost from radicchio leaves during processing was determined by comparing E. coli O157: H7 populations on the radicchio samples before (control) and after washing using direct plating with or without membrane filtration, depending on the inoculation level. E. coli O157:H7 populations were also assessed in two 50 ml wash water samples; one was taken at the beginning as a control and 28

another one was taken at the end of washing. One centrifugation water sample was also assessed for E. coli O157:H7 using membrane filtration. 2.2.2 Produce. Individually wrapped romaine lettuce (12 heads/case) and radicchio heads (9 heads/case) were obtained from a local wholesaler (Stan Setas Produce Co., Lansing, MI), stored in a 4 o C walk-in cooler, and used within 3 days of delivery. On the day of use, the outer romaine lettuce leaves were discarded with the remaining leaves separated until the core was reached. For the processing experiments, one 5 kg batch of romaine lettuce was mechanically shredded using the Urschel shredder for each processing run, and then collected for flume washing. 2.2.3 Bacterial strains used. For safety purposes, all of this work was conducted using non-toxigenic, avirulent strains. Recent studies of bacterial attachment ability were conducted in our laboratory with a modification of the microtiter plate assay described by Jackson t al. (2002). Based on the results, these avirulent strains behaved similarly to a set of virulent strains linked to outbreaks involving these same commodities (Buchholz et al., 2012; Wang and Ryser, 2014). Four non-toxigenic, green fluorescent protein labeled (GFP-labeled), ampicillin-resistant strains of E. coli O157:H7 (ATCC 43888, CV2b7, 6980-2, and 6982-2) previously obtained from Dr. Michael Doyle (Center for Food Safety, University of Georgia, Griffin, GA) were used and prepared in trypticase soy broth (Difco, Becton Dickinson, Sparks, MD) containing 0.6% yeast extract (Difco, Becton Dickinson) and 100 ppm ampicillin (ampicillin sodium salt, Sigma 29

Chemical Co., St. Louis, MO) (TSBYE-AMP) and 10% (v/v) glycerol (Sigma Chemical Co., St. Louis, MO) and stored at -80 C until use. Working cultures were prepared by streaking each stock culture on trypticase soy agar plates (Difco, Becton Dickinson) containing 0.6% yeast extract and 100 ppm ampicillin (TSAYE-Amp). After 18 24 h of incubation at 37 C, a single colony was subjected to two successive transfers in 9 ml of TSBYE-Amp and incubated at 37 o C for 24 h. After incubation, the cultures were combined in equal volumes to obtain a 4-strain cocktail containing ~10 9 CFU/ml, and then diluted in sterile phosphate buffer solution (PBS) to the levels needed for inoculation. 2.2.4 Inoculation of radicchio. In order to assess redistribution of the inoculum to previously uncontaminated product, the required quantities of radicchio (aseptically cut into pieces measuring ~ 5 x 5 cm with an average weight of 1 g per piece) were spot-inoculated with a total of 50 µl of the E. coli O157:H7 cocktail at multiple locations (normally 5-8 locations) on the leaf to obtain populations of 10-1, 10 1 or 10 3 CFU/leaf after overnight storage at 4 o C. After 15 minutes of drying in a biosafety cabinet, these inoculated surrogate products were collected into sterile plastic boxes (PLA NatureWorks, NE) and stored overnight at 4 o C to allow the inoculum to attach to the leaf surface to mimic pre-harvest contamination. Just before processing, a 25 g radicchio sample was assessed to determine the starting inoculation level. Four different inoculated:uninoculated product ratios (0.5:100, 1:100, 5:100, 10:100) based on product weight were used to determine the differences in pathogen 30

redistribution based on the amount of inoculated to uninoculated product processed for each experiment.. 2.2.5 Processing line. A pilot-scale processing line consisting of a lettuce shredder, step conveyer, flume tank, shaker table, and dewatering centrifuge was used for all experiments. The commercial lettuce shredder (model TRS 2500 Urschel TranSlicer, Valparaiso, IN) was operated at a feed belt/slicing wheel speed of 198 m/min and 905 RPM, respectively, to obtain a shred size of approximately 5 5 cm. A stainless steel non-refrigerated water recirculation tank (~1000 L capacity) was connected to a 3.6-m long stainless steel flume tank (Heinzen Manufacturing, Inc., Gilroy, CA) equipped with two overhead spray jets by a 4.1-m long, 10-cm diameter hard plastic discharge hose and a centrifugal pump (model XB754FHA, Sterling Electric, Inc., Irvine, CA) that circulated the water at ~15 L/sec. A custom-made stainless steel screen was installed at the end of the flume tank to retain product for 90 seconds of washing. The stainless steel shaker table for partial dewatering was operated by a 1 HP Baldor washdown duty motor (Baldor Electric Co., Ft. Smith, AR) at 1760 RPM. Water removed from the leafy greens during mechanical shaking passed through a fine mesh screen and was fed into the water holding tank by a water recirculation spout underneath the shaker table. A 22.7-kg capacity centrifugal Spin Dryer (model SD50-LT, Heinzen Manufacturing, Inc.) with three internally timed spin cycles totaling 60 s was used for centrifugal drying. The water used for this processing line was tap water at ~10 o C. 31

2.2.6 Romaine lettuce processing and sample collection. Uninoculated romaine lettuce leaves (5 kg) were hand-fed into the shredder (Figure 2.1) at a rate of approximately 0.5 kg/second, and mixed with the required amount of inoculated radicchio (0.5:100, 1:100, 5:100, 10:100 on a per weight basis). Thereafter, the leaves were manually dumped into the flume tank (Figure 2.2), washed in 890 L of recirculating sanitizer-free water (~10 o C) for 90 s, released from the flume tank, partially dewatered on the shaker table, collected in a single centrifugation basket and then centrifugally dried (Figure 2.3). After removing all the inoculated radicchio for separate testing, the entire 5 kg batch of romaine lettuce was subdivided into 225 g samples in separate Whirl-pak bags for further analysis with the spent centrifugation water (Figure 2.4). A 25 g sample of inoculated radicchio was removed from the total removed and examined for numbers of E. coli O157:H7 after processing. One 50 ml wash water sample was collected from the water return spout above the recirculation tank before processing (control) and one 50 ml wash water sample was collected at the end of processing and tested for E. coli O157:H7. Finally, one 225g romaine lettuce sample per delivery was tested upon arrival using GeneQuence kit to ensure absence of E. coli O157:H7. 32

Figure 2.1: Lettuce shredding Figure 2.2: Flume washing and dewatering 33

Figure 2.3: Centrifugation Figure 2.4: Sample collection 34

2.2.7 Chemical sanitizers. To mimic industrial practices more closely, romaine lettuce was also processed as above using flume water containing 60 ppm of available chlorine (XY-12, Ecolab, acidified to ph 6.50 with citric acid, Sigma-Aldrich, St. Louis, MO) (Davidson et al., 2013). However, considering that E. coli O157:H7 was expected to be non-detectable after processing, only the highest product inoculation level (10 3 CFU/leaf) and inoculated: uninoculated ratio (10:100) were assessed for romaine lettuce based on three replicates. 2.2.8 Microbiological analysis. All of the samples including 25 g of radicchio before processing (control), 25 g of radicchio after processing, three water samples and the ~20 225 g bags of romaine lettuce were qualitatively and quantitatively assessed for E. coli O157:H7 using direct plating and the GeneQuence assays (Neogen Corp., Lansing, MI). The 25 g radicchio samples were diluted in 50 ml sterile PBS and homogenized using a stomacher (Stomacher 400 Circulator, Seward, Worthington, UK) at 260 rpm for 2 min. Thereafter, 50 ml of the sample homogenate or 50 ml of wash and centrifugation water collected after processing were plated with or without prior membrane filtration on TSAYE-amp to quantify E. coli O157:H7 after 24 h of incubation at 37 o C. The lettuce sample homogenate was enriched using the Neogen Reveal 20 h enrichment method (Neogen Corp., Lansing, MI) and examined for the presence/absence of E. coli O157:H7 using the GeneQuence assay according to the manufacturer. Due to the technical limitation in quantification of low levels of contamination, E. coli O157:H7 populations in each lettuce sample bag after processing were not assessed. 35

2.2.9 GeneQuence Assay. The GeneQuence assay is a sandwich enzyme-linked immunosorbent assay (S-ELISA) in a microwell format, which can be used to screen various commodities for the presence of E. coli O157:H7 antigens (Neogen, 2016). The GeneQuence assay contains capture and detector DNA probes specific to the ribosomal RNA of E. coli O157:H7. A positive result occurs when both DNA probes bind to the target and the capture probe-coated well (Neogen, 2016). The instruments used for GeneQuence testing include a Stat Fax 4200 microplate reader and a Stat Fax 2600 automatic microplate washer provided by Neogen (Figure 2.5). The GeneQuence kits (Figure 2.6) were stored refrigerated in a 4 o C walk-in cooler and equilibrated to room temperature before use (Neogen, 2016). 36

Figure 2.5: The Stat Fax 4200 microplate reader and Stat Fax 2600 semi-automatic microplate washer Figure 2.6: GeneQuence kit 37

2.2.10 Statistical analysis. The quantitative results obtained from direct plating were converted to log CFU/g or log CFU/100 ml to calculate percent transfer based on the initial inoculum. In addition, the percentage of E. coli O157:H7 CFU lost during processing was determined from the inoculated product counts before and after processing. The qualitative results were analyzed by determining the percentage of bags positive for E. coli O157:H7 based on the total weight of product processed. Results based on triplicate experiments were averaged and subjected to an analysis of variance (ANOVA) using JMP 12.2 (SAS Institute Inc., Cary, NC). The Tukey- Kramer HSD test was used with P values of 0.05 considered significantly different. Values of half the limit of detection were used when no E. coli O157:H7 were detected in the samples. 38

2.3 RESULTS 2.3.1 Romaine lettuce. After collecting and enriching samples overnight, GeneQuence testing was performed to determine the presence/absence of E. coli O157:H7 in the sample bags. Based on triplicate experiments, all lettuce samples yielded E. coli O157:H7 when radicchio was inoculated at 10 3 CFU/leaf. When radicchio was inoculated at 10 1 CFU/leaf, the percentage of positive samples decreased from 96.8% to 93.7%, 81.0% and 63.5% for inoculated: uninoculated ratios of 10:100, 5:100, 1:100 and 0.5:100, respectively. At 10-1 CFU/leaf, 22.2%, 6.3%, 4.8% and 6.3% of the samples were positive for the same inoculated:uninoculated ratios (Table 2.1). Based on triplicate experiments, when the inoculation level decreased from 10 3 CFU/leaf to 10-1 CFU/leaf, the populations of E. coli O157:H7 transferred from radicchio to romaine lettuce decreased as well (P < 0.05). Within each inoculation level, no significant differences (P > 0.05) were found among the four inoculated: uninoculated product ratios. The results are presented as graphs in Figures 2.7 and 2.8 below. 39

Figure 2.7: Percent of samples detected as positive for E. coli O157:H7 due to the E. coli O157:H7 transfer to 225 g samples of romaine lettuce after 90 sec of flume washing in sanitizer-free water based on different inoculation levels. Mean values with different letters are significantly different (P 0.05) within the same product ratio (10:100, 5:100, 1:100, 0.5:100). 40