Introduction

An estimated 9 million illnesses are due to known foodborne pathogens every year (Scallan et al., 2011). Between 1998 and 2008, contaminated produce contributed to 46% of illnesses with leafy greens associated with 22% of the illnesses alone (Painter et al., 2013). Produce is often attributed to foodborne pathogen transmission because it is consumed raw unlike other food products that include a pathogen inactivation step, such as pasteurization, cooking, or sterilization (Benson, 2010). Pathogens that are often associated with produce outbreaks include Escherichia coli O157:H7, Salmonella enterica subsp. enterica, Listeria monocytogenes, and human norovirus (HuNoV) (Grant et al., 2008; Greene et al., 2008; Hall et al., 2012).

Contamination of produce can occur at any point in the food supply chain. Pathogens can be transferred to produce from water, soil, animal excretions, contaminated surfaces, or during handling and preparation in retail foodservice operations or by the consumer (Benson, 2010). Retail settings present several opportunities for cross-contamination of fresh produce, including the mishandling of the product, inadequate handwashing, and poor hygiene by employees. Aside from the implementation of good handling practices (GHP), wash treatments (e.g., chlorine, peroxyacetic acid, acidified sodium chlorite) are the primary prevention method utilized for fresh produce at retail (Olaimat & Holley, 2012). However, alternate approaches for the inactivation of pathogens on fresh produce in retail foodservice settings exist. One such approach includes the addition of ozone—a strong oxidant and sanitizer—to wash water used for fresh produce prior to preparation (Rosenblum et al., 2012).

By applying an ozone wash into the retail foodservice setting, this additional control step could inactivate pathogens possibly present on the produce and reduce spoilage organisms, thus extending shelf life. When using ozone as a sanitizer, several variables impact efficacy. Extrinsic factors, such as water temperature, free chlorine (i.e., present in municipal tap water), and pH, affect the stability of the ozone and ultimately the efficacy (Horvitz & Cantalejo, 2014). Besides extrinsic factors, intrinsic factors need to be considered: concentration and form (gaseous or aqueous) (Kim et al., 2003). Ozone has been shown to inactivate viruses (e.g., murine norovirus, poliovirus, and human rotavirus) and bacteria (e.g., Escherichia coli O157:H7, Listeria monocytogenes, and Pseudomonas fluorescens) while in suspension (Kim & Yousef, 2000; Lim et al., 2010; Roy et al., 1982; Vaughn et al., 1987). However, while studies have been published on the efficacy of ozone against bacteria on fresh produce, there is very limited research on the reduction of viruses on these same matrices.

Another factor to consider are microbe-virus interactions on fresh produce surfaces and the potential impact on sanitizer efficacy (Dawley & Gibson, 2019; Deng & Gibson, 2017). It has been hypothesized that these virus-bacteria interactions may allow microorganisms, specifically enteric viruses, to resist current control strategies and place the consumer at an increased risk for foodborne illness. Murine norovirus and poliovirus (serotype 1, Mahoney) have been shown to interact with Gram-positive and Gram-negative bacteria, including Bacillus cereus, Enterococcus faecalis, E. coli, and Enterobacter cloacae (Deng et al., 2019; Jones et al., 2014; Kuss et al., 2011). Specifically, this study is interested in B. cereus and E. cloacae.

Bacillus cereus is a Gram-positive bacterium that is found in the soil and has also been reported on produce (Stenfors Arnesen et al., 2008). E. cloacae is a Gram-negative organism that is found in the environment and the intestinal tract of humans and is often linked to nosocomial infections (Harbarth et al., 1999). Similar to B. cereus, E. cloacae has also been found on the surface of produce (Al-Kharousi et al., 2016). Since these bacteria naturally occur on the surface of produce, the bacteria could possibly interact with viruses introduced to the produce surface (Deng & Gibson, 2017; Moore & Jaykus, 2018) and potentially alter the ability of sanitizers to inactivate viruses.

In the limited number of studies, it has been shown that when viruses interact with bacteria expressing specific surface antigens, such as histo-blood group antigens (HBGA), viruses may be more resistant to disinfectant processes. For example, HuNoV-like particles associated with HBGA-expressing bacteria displayed greater capsid integrity and binding capabilities than HuNoV-like particles alone following heat treatment at 90 °C for 2 min (Li et al. 2015). Conversely, Li et al. (2017) published results indicating that Tulane virus—a cultivable HuNoV surrogate—was not protected from heat while interacting with bacteria expressing HBGA-like molecules. Meanwhile, Deng et al. (2019) also reported a lack of protection against thermal and chemical inactivation for Tulane virus, murine norovirus, and Aichi virus when in association with E. cloacae. These conflicting results raise several questions when it comes to the interactions of viruses and bacteria, and how these interactions may impact virus inactivation under specific sanitation practices. Therefore, this study aimed to evaluate the efficacy of aqueous ozone on the inactivation of (i) viruses, (ii) bacteria, and (iii) viruses in association with bacteria on the surface of leaf lettuce and tomatoes.

Materials and Methods

Propagation of Virus Surrogates

MS2 stock was generated as described previously by Gibson et al. (2012). MS2 was kindly provided by Dr. Stephanie Friedman from the United States Environmental Protection Agency (EPA) Gulf Ecology Division in Gulf Breeze, FL. Stocks of MS2 were generated utilizing E. coli C3000 bacterial host (ATCC 15597; American Type Culture Collection, Manassas, VA) as previously described by Gibson et al. (2012) with slight modifications. Briefly, viral cell lysate was suspended in 23 ml of 1 × phosphate buffered saline (PBS), vortexed, and centrifuged at 185×g for 25 min. The supernatant was collected and filtered through a 0.22 µm filter (Millipore Corporation, Billerica, MA), aliquoted, and stored at − 80 °C. The MS2 stock concentration was determined based on titration by double agar layer (DAL) method with bacterial host E. coli C3000.

Murine norovirus type 1 (MNV) was prepared as described previously by Deng et al. (2019) with modifications. MNV was kindly provided by Dr. Kellogg Schwab (Johns Hopkins Bloomberg School of Public Health, Baltimore, MD). MNV was propagated in monolayers of RAW 264.7 (mouse leukameic monocyte macrophage, ATCC TIB-71) cells with all steps conducted at 37 °C with 5% CO2. Cells were cultured in Dulbecco modified Eagle’s medium (DMEM) (Sigma-Aldrich, St. Louis, MO) containing 10% low endotoxin, fetal bovine serum (FBS: GibcoLife Technology, Gaithersburg, MD), 1% 100 × penicillin–streptomycin solution (GibcoLife Technology), 1% HEPES (Sigma-Aldrich), 1% glutamine (Hyclone, Logan, UT), and 1% non-essential amino acids (Corning, New York, NY). After reaching 90% confluence, cells were infected with MNV at a multiplicity of infection (MOI) of 0.05 for virus stock production. The virus was extracted from cell lysate after complete cytopathic effect (Deng et al. 2019).

For MNV titration, 6-well plates were seeded with 2 × 106 cells per well and grown to 90% confluence in 2 ml of complete growth medium. Cell monolayers were inoculated with 500 µl serial diluted virus stock for 1 h with rocking followed by removal of the inocula. Cells were covered with 2 ml of prepared overlay medium containing: 25% of 6% low melting point agarose, 50% 2 × minimum essential medium eagle (MEM) (Corning), 10% low endotoxin FBS, 1% 100 × penicillin–streptomycin solution, 1% glutamine, 0.5% HEPES, and 12.5% sterile-distilled water. The plates were incubated for 72 h. Next, 2 ml of 0.01% neutral red (Sigma-Aldrich) prepared in 1 × PBS was added to each well to visualize plaques. Plaque forming units (PFU) were counted after 1 h to determine virus titer (PFU/ml).

Cultivation of Bacteria

Enterobacter cloacae (ATCC 39979) with rifampicin resistance was streaked from a frozen 50% glycerol stock onto Luria Bertani (LB) agar (Alfa Aesar, Tewksbury, MA) with 100 µg/ml of rifampicin (Alfa Aesar) using a sterile inoculation loop and incubated overnight at 37 °C. A single CFU was selected from the plate, placed in 5 ml of LB broth with 100 µg/ml rifampicin in a 50 ml centrifuge tube, and incubated overnight at 37 °C with shaking at 150 rpm. Bacillus cereus (ATCC 14579) was also streaked from a frozen, 50% glycerol stock onto a nutrient agar (NA) plate (Becton, Dickinson and Company, Franklin Lakes, NJ) using a sterile inoculation loop and incubated overnight at 30 °C. A single CFU was selected from the plate, placed in 5 ml of nutrient broth (NB) (Becton, Dickinson and Company) in a 50 ml centrifuge tube, and incubated overnight at 30 °C with shaking at 150 rpm.

Following overnight growth of both bacteria, culture tubes were centrifuged at 5000×g for 10 min to pellet the bacterial cells. The supernatant was decanted, and the pellet was resuspended in 5 ml of 1 × PBS, vortexed, and centrifuged again. The bacterial pellet was resuspended in 5 ml of 1 × PBS, and the concentration of each bacterial culture was determined by spread plate enumeration of microorganisms. LB agar with 100 µg/ml of rifampicin and B. cereus agar (Oxoid, Altrincham, Cheshire, England) with selective supplement (Oxoid) and egg yolk emulsion (Dalynn, Calgary, Canada) were used for E. cloacae and B. cereus, respectively.

MNV-Bacteria Interaction

Enterobacter cloacae and B. cereus were serially diluted to 106 CFU/ml in 1 × PBS. For interactions, 100 µl of bacteria were mixed with 100 µl of MNV (106 PFU/ml) at room temperature in a microcentrifuge tube and allowed to associate for 1 h. This process was confirmed previously by Almeida and Gibson (2017) to result in the complete association of MNV with both E. cloacae and B. cereus. The assumption of complete association is based on an assay detection limit of 2 PFU/ml and no detection of the virus after passing the sample through a 0.22 µm syringe filter. However, it is important to note that virus aggregates could be present and may not pass through the 0.2 µm filter, and thus, complete association cannot be confirmed with absolute confidence (Samandoulgou et al. 2015).

Types of Produce

Boston bibb (BB) loose-leaf lettuce and cherry tomatoes were used. The BB was grown hydroponically and procured from two separate places: (1) grown in a greenhouse at Ozark All Seasons in Windsor, AR, and purchased from a local grocer and (2) grown in a freight farm—a hydroponic farming system that is built into a shipping box—on the University of Arkansas Fayetteville campus and donated. Upon harvest, BB leaves were placed in a sterile plastic container with a lid and stored at 4 °C until use. Cherry tomatoes were received from a local produce distributor or from a local grocery store and stored at 4 °C until use. No pre-treatment was applied to the produce to remove associated microorganisms prior to inoculation of produce with viruses and bacteria used in the present study.

Inoculation of Produce

For virus inoculation, 25 g of BB leaves were allocated for each sampling time point (n = 5). Each 25 g portion was spot inoculated with 100 µl of MS2 (108 PFU/ml) and 100 µl of MNV (106 PFU/ml). For cherry tomatoes, 2 tomatoes (20–25 g total) were included for each sampling point (n = 5) and inoculated similarly to BB. For bacteria, the same amount of produce and inoculation methods were used as described for viruses. The BB was inoculated with 100 µl of 107 CFU/ml E. cloacae and 100 µl of 106 CFU/ml B. cereus. Due to a low recovery of bacteria based on preliminary work, tomatoes were inoculated with 100 µl each of 108 CFU/ml E. cloacae and B. cereus. For interaction experiments, 25 g of BB was spot inoculated with the mixture containing MNV associated with bacteria (see “MNV-Bacteria Interaction”).

Following inoculation, the produce was allowed to sit at room temperature until the surface was visibly dry where inoculated. For virus experiments, the produce was stored overnight (24 h) or for 90 min at 4 °C and then exposed to the ozone wash or to the water-only wash. For bacteria experiments, the produce was stored overnight (24 h) at 4 °C prior to the ozone wash or to the water-only wash. Based on results, following association of MNV with bacteria (see “MNV-Bacteria Interaction”), the prepared inocula were allowed to associate with the produce for 90 min at 4 °C prior to treatment.

Wash Treatments

The batch wash ozone sanitation system (BWOSS) utilized in the present study was developed by Recycled Hydro Solutions (Rogers, AR) under the trade name RinseWell®. The unit fabricated for this study is comprised of a single compartment sink measuring 43 cm2 and 30 cm deep. The sink holds approximately 34.07 l (9 gallons) of water. During operation, the sink is filled with water, and then the water is passed through a Venturi injector, thus creating ozone continuously. The ozonated water is then recirculated back into the sink basin. Ice was added to the sink to aid in stabilization of ozone until the water temperature reached 4 °C. To determine ozone concentrations, the BWOSS was fitted with a dissolved ozone meter (Model Q46, ATI, Collegeview, PA) and ozone readings were corroborated by the indigo trisulfonate method (SM 4500-OS3 B) using a Hach Pocket Colorimeter II (Hach Company, Loveland, CO) and Ozone AccuVac Ampules (Hach) (American Public Health Association, American Water Works Association, and Water and Environment Federation 2012). The produce was placed in the BWOSS once ozone concentrations reached 0.5 ppm—as indicated by the dissolved ozone meter—and the water temperature in the sink was stable at 4 °C. Prepared produce samples were submerged in the sink with subsamples (25 g portions of BB or 2 cherry tomatoes) taken every 10 min for a total of 40 min. Time zero samples (i.e., no wash) were also collected.

For this study, we chose to only control ozone concentrations with water temperature and allowed the system to operate as intended, injecting ozone continuously with typical fluctuations in ozone concentrations over time. The system is designed to achieve and maintain ozone concentrations of approximately 0.8 to 1 ppm under optimal conditions.

Recovery and Detection of Microorganisms

Following treatment, produce samples were placed in Whirl–pak bags with 75 ml of buffered phosphate water (BPW). The bags containing lettuce were then placed in a stomacher (Seward Stomacher 400 Circulator, West Sussex, United Kingdom) for 1 min at 260 rpm. The eluate was then serially diluted. The cherry tomatoes were hand massaged for 1 min instead of stomaching to prevent the tomatoes from breaking. Based on preliminary data, we speculate that the pH of the eluate dropped when the tomatoes broke resulting in interference with the MNV plaque assay due to cytopathic effects on the RAW 264.7 cells. The resulting eluate from the hand-massaged tomatoes was then serially diluted in BPW.

For MS2, 100 μl of each dilution was plated in duplicate using the DAL method. Then the PFUs were counted, and PFU/ml were calculated. For MNV, plaque assay was performed as previously described (Sect. 2.1.2), and PFU were counted to calculate PFU/ml. Virus experiments were conducted at least in duplicate. For bacteria, the eluate was ten-fold serially diluted, and 100 μl of each dilution was plated on LB agar with 100 μg/ml of rifampicin and B. cereus agar for detection of E. cloacae and B. cereus, respectively. The plates were incubated overnight at 37 °C and 35 °C, respectively. Colonies were counted to determine CFU/ml. All bacterial experiments were conducted in duplicate.

MNV from interaction experiments was recovered from leaf lettuce by washing with 75 ml of BPW containing 40 µg/ml of penicillin/streptomycin to prevent bacteria interference with the plaque assay. The bags were then placed in the stomacher for 1 min at 260 rpm, or hand massaged. A plaque assay was performed as previously described in ‘Propagation of Virus Surrogates.’ PFUs were counted, and then PFU/ml were calculated.

Data Analysis

Data analyses were conducted in R version 3.3.2 (R Core Team, Vienne, Austria). Both Kruskal–Wallis (KW) test and one-way analysis of variance (ANOVA) F test were conducted to identify differences among the three treatments for inactivation of virus alone: Ozone (24 h), Ozone (90 min), and Water. These tests revealed similar results. Thus, Kruskal–Wallis test results are reported at a significance of P ≤ 0.1. For inactivation of bacteria and MNV + bacteria on produce, ANOVA t test was selected over KW test since the t test can provide information about the effect size and is more sensitive to the numerical scale of our response variable. For the t-test, the assumption of normality was checked using the Shapiro–Wilk test and resulted in no concern. A significance level of P ≤ 0.1 was also applied in this instance.

Results

Treatment conditions

The average ozone concentration for each treatment is shown in Table 1. The ozone concentration ranges from 0.48 to 0.99 ppm. The treatments all have an initial concentration of at least 0.48 ppm of ozone. As time increases, the ozone concentration also increases, reaching at least 0.85 ppm after 40 min for most treatments. Water temperature was initially 3 to 4 °C and over the 40 min increased to 7 to 8 °C. The pH was consistent with an average of 7.6 across all treatments.

Table 1 Mean ozone concentrations over time (min) for all treatments
Full size table

Virus Inactivation

A baseline recovery was used to calculate the amount of virus that can be recovered from each produce without any treatment (Fig. S1). After a 24 h attachment period, the average recovery from BB was 50.1% (1.86 × 107 PFU) and 93.2% (1.15 × 105 PFU) for MS2 and MNV, respectively. Meanwhile, the average recovery of MS2 and MNV from cherry tomatoes was 37.5% (1.39 × 107 PFU) and > 100% (5.44 × 105 PFU), respectively. The treated samples were then compared to the baseline recovery at 0 min for each experiment to give the log reductions over time for the various treatments.

With the exception of the 10 min treatment, MS2 inactivation on BB did not reveal a significant difference between the ozone wash and the water-only wash or between the two attachment times (Fig. 1). When comparing ozone washes to water-only washes, there was a greater log reduction based on PFU/ml values with ozone resulting in a greater reduction of MS2 at both attachment times; however, there was no significant difference. Similarly, MNV inactivation on BB did not reveal a significant difference between the ozone wash and the water-only wash or between the two attachment times (Fig. 1). When comparing ozone washes to water-only washes, there was a generally a greater reduction with ozone than water though this observed trend was not statistically significant.

Fig. 1
figure 1

Mean log reduction for viral surrogates on Boston bibb. Log reductions from three treatments are displayed below with dots within MNV and MS2 bacteriophage, respectively, over time. Here, 90 min means that the virus was inoculated onto the produce, dried at RT, and then stored in 4 °C for 90 min. Similarly, 24 h means that the virus was inoculated onto the produce, dried at RT, and then stored in 4 °C for 24 h. Kruskal–Wallis test was performed in each of eight situations and the P value is reported when any difference between treatments is revealed at 0.1 level. The limit of detection for MNV and MS2 was 10 PFU/ml in a 75 ml sample

Full size image

In Fig. 2, MS2 and MNV inactivation on cherry tomatoes are reported. For MNV, unlike the BB results (Fig. 1), the 24 h attachment on tomatoes had a greater reduction than the 90 min attachment when treated with aqueous ozone (Fig. 2). There was also variability in the log reduction trend of MNV over time when comparing ozone and water-only treatments.

Fig. 2
figure 2

Mean log reduction for viral surrogates on cherry tomatoes. Log reductions from three treatments are displayed below with dots within MNV and MS2 bacteriophage, respectively, over time. Here, 90 min means that the virus was inoculated onto the produce, dried at RT, and then stored in 4 °C for 90 min. Similarly, 24 h means that the virus was inoculated onto the produce, dried at RT, and then stored in 4 °C for 24 h. Kruskal–Wallis test was performed in each of eight situations and none of the tests revealed the significant difference between treatments at 0.1 level. The limit of detection for MNV and MS2 was 10 PFU/ml in a 75 ml sample

Full size image

Bacterial Inactivation

Baseline recoveries for bacteria were also used to calculate the log reduction over time (Fig. S1). The average recovery of E. cloacae from BB and cherry tomatoes was 31.4% (1.53 × 106 CFU) and 13.2% (8.89 × 106 CFU), and the average recovery of B. cereus from BB and cherry tomatoes was > 100% (2.74 × 105 CFU) and 5.24% (8.18 × 105 CFU), respectively. Bacterial inactivation on BB is shown in Fig. 3a, and ozone generally achieved greater reduction in bacteria than water-only. However, the only statistically significant differences were observed for B. cereus and E. cloacae at 20- and 10-min exposure times, respectively. Ozone treatment of cherry tomatoes had a higher observed log reduction than water-only for both bacteria with statistically significant differences recorded at 10- and 20-min for B. cereus (Fig. 3b). When comparing the two bacteria on BB, E. cloacae was reduced by 3.33 log CFU/ml after a 40 min ozone treatment compared to 2.83 log CFU/ml for B. cereus. Conversely, B. cereus was reduced by 2.82 log CFU/ml compared to 2.58 log CFU/ml for E. cloacae on cherry tomatoes after 40 min exposure to ozone wash.

Fig. 3
figure 3

Mean log reduction of bacteria on produce. a Log reduction on Boston bibb, b Log reduction on tomatoes from two treatments are displayed below with dots within B. cereus and E. cloacae, respectively, over time. The t-test was performed, reporting the P value if the difference between treatments revealed at 0.1 level. The limit of detection for bacteria was 10 PFU/ml in a 75 ml sample

Full size image

Inactivation of MNV in Association with Bacteria

In Fig. 4, the results from the ozone and water-only wash treatments are displayed. As described previously, baseline recovery of virus from BB after 24 h attachment was used to calculate the log reduction for treatment times. Results for MNV, as shown in Fig. 1, are used in comparison to the virus-bacteria interactions. Except for the 40 min ozone wash treatment, there were no significant differences between MNV only and any of the virus–bacteria interactions. More specifically, after 40 min exposure to ozone (maximum ozone level of 0.86 ppm, refer to Table 1), the log reduction for MNV, MNV with B. cereus, and MNV with E. cloacae were 1.63, 2.60, 1.83 PFU/ml, respectively (P = 0.009). Overall, MNV and B. cereus interaction had consistently greater log reductions over time compared to the MNV with E. cloacae and MNV only log reductions at 30 and 40 min.

Fig. 4
figure 4

Mean log reduction for interaction experiments. Log reductions for MNV-bacteria interaction are displayed below with dots in each time point by treatment. Kruskal–Wallis test was performed in each of eight situations and none of the tests revealed the significant difference between the interactions at 0.1 level. ANOVA F test also was performed, and the P value was reported at < 0:1. The limit of detection for MNV was 10 PFU/ml in a 75 ml sample

Full size image

Discussion

A primary conclusion from this study was that ozone did not consistently provide a statistically significant impact on virus reduction on BB or cherry tomatoes when compared to water-only treatments under the conditions tested. Regardless, average log reductions of 2.40 (± 0.97) and 2.97 (± 1.16) for viruses on BB and tomatoes, respectively, were still achieved when exposed to aqueous ozone across all treatment time points and attachment conditions. When examining the effects of aqueous ozone on viral inactivation, very few studies have investigated the inactivation of viruses on fresh produce.

Hirneisen et al. (2011) reported a more than three-log reduction of MNV on lettuce and green onions after a 10 min exposure to 6.25 ppm ozone. Moreover, the authors observed a five-log reduction of MNV inoculated in water after a 10 min exposure to ozone and concluded that the food matrices impacted viral inactivation rates. However, Hirneisen and co-authors did not publish the results of a water-only wash, but rather referenced that a water wash was performed as a control. By comparing ozone and water-only treatments in the present study, these results contradict those reported by Hirneisen et al. (2011). We report that ozone, in general, did not have a statistically significant impact on log reduction of viruses. It may be hypothesized that virus reduction on produce is primarily due to physical removal followed by inactivation of viruses in the ozonated wash water. However, these discrepancies could be due to the differing produce sample sizes, the volume of ozonated water the samples were treated with, the amount of organic load, or the applied ozone concentration. More specifically, Hirneisen et al. (2011) treated 5 g of produce in a flask containing 45 ml of sterile water with bubbling gaseous ozone to achieve a concentration of 6.25 ppm. Conversely, in the current study, 25 g of produce was submerged in a sink with 34 L of ozonated water at a maximum concentration between 0.74 ppm and 1 ppm.

Organic load is also an important driver in the efficacy of ozonated water—i.e., excess organic materials can decrease ozone efficacy (Gibson et al. 2019). However, aside from the produce itself, there was no ‘added’ organic load in the present study nor as reported by Hirneisen et al. (2011). Even though there are conflicting results, both studies indicate that food matrices most likely protect viruses from ozone inactivation. Due to the variability in the results presented here, stainless steel coupons were used to evaluate the efficacy of aqueous ozone against the viruses without the food matrix. The removal of the food matrix led to a decrease in variability, increase in log reduction over time, and a visible difference between ozonated water and water alone (Fig. S2).

In the present study, it was also shown that attachment time for viruses did not affect the efficacy of ozone. To our knowledge, no other published studies have investigated the attachment time of viruses to fresh produce and the impact on sanitizer efficacy. The results reported here show that the attachment time of viruses has no significant effect with respect to ozone inactivation.

Even though the results are similar for both BB and cherry tomatoes, there were visible differences in the reductions of viruses and bacteria between the two produce types. The difference in reduction could be due to the surface topography of the produce. Lu et al. (2015) investigated the influence of the epicuticular surfaces of lettuce and tomatoes as it related to the adsorption of rotavirus. The authors found that these surfaces play a role in the effectiveness of sanitation treatments and could cause differences in virus log reductions between produce types. Interactions of the microorganism with the produce may also play a role in the observed difference in microbial inactivation between produce types. A majority of HuNoV are known to bind to HBGA on cells lining the gastrointestinal tract, which can lead to infections in humans (Ettayebi et al., 2016). It has been previously observed that lettuce possesses HBGA-like carbohydrates and other carbohydrate moieties that norovirus particles could attach to and thus could not be removed by simple washing (Gao et al., 2016). The HuNoV surrogate used in the present study, MNV, binds to glycans featuring terminal sialic acid moieties and cell surface glycoproteins (i.e., CD300lf) required for viral internalization (Taube et al., 2009; Orchard et al., 2016). These types of interactions could be an explanation as to why there was a lower reduction of viruses on lettuce compared to tomatoes.

With respect to bacteria, Takeuchi et al. (2000) reported that genera of bacteria (E. coli O157:H7, L. monocytogenes, Salmonella Typhimurium, and Pseudomonas fluorescens) attach to lettuce differently. The authors observed that E. coli and L. monocytogenes attached to cut edges while P. fluorescens preferred the surface of the lettuce; meanwhile, Salmonella did not have a preference in attachment site. Takeuchi and co-authors further described that these differences are due to the ability of the bacteria to bind to the hydrophobic cuticle layer. The bacterial binding could explain why log reductions are different between E. cloacae and B. cereus for the produce analyzed in the present study. In general, the data demonstrate that wash water with ozone does not significantly impact the removal and/or inactivation of E. cloacae and B. cereus on either produce type.

Previous studies on ozone inactivation of bacteria on fresh produce report findings that conflict with those reported here. Kim et al. (1999a) investigated the inactivation of P. fluorescens on the surface of shredded lettuce. The authors found that bubbling ozone into the wash water was significantly different from the water-only wash for inactivation of P. fluorescens. However, their system design did not resemble a practical, real-world application as samples were placed in a beaker with 500 ml of water and stirred while the ozone was generated. Moreover, the ozone concentration in Kim et al. (1999a) reached 10 ppm—ten-fold greater than the present study. Similarly, Selma et al. (2007) concluded that ozonated wash water significantly reduced Shigella sonnei inoculated on shredded lettuce. This experiment took place in a 50 l tank, and the sample size was 30 g, with the longest exposure time being 5 min at various ozone concentrations (1, 2, and 5 ppm). In the present study, the sample size was 25 g, which is similar to Selma et al. (2007), but the studies differ in contact times and in the concentrations of ozone applied.

The bacteria investigated here were chosen because both bacteria can be present in the phyllosphere of leafy greens and fresh produce in general (Al-Kharousi et al., 2016; Stenfors Arnesen et al., 2008). In addition, these bacteria also represent both Gram-positive and Gram-negative species. Gram-negative bacteria are reportedly more sensitive to ozone due to the composition of the cell membrane not containing as much peptidoglycan compared with Gram-positive bacteria (Kim et al., 1999a, 1999b). Although that was true for Gram-negative bacteria on BB in the present study, ozone had a greater reduction of Gram-positive bacteria on tomatoes. This could mean that produce type is a crucial aspect in the inactivation of the microorganisms. It is also important to note that unpublished data has shown that when the BWOSS is drained, detectable levels of microorganisms were left on the surface of the sink with water-only washes, whereas there were no microorganisms detected on the surface of the sink when ozone was used. These differences indicate that adding ozone to wash water could prevent subsequent cross-contamination during produce washing.

The research presented here is the first to investigate the effects of aqueous ozone on the inactivation of a norovirus surrogate while in combination with bacteria. Studies on interactions of HuNoV and its surrogates with bacteria have increased dramatically over the past decade (Dawley & Gibson, 2019; Amarasiri & Sano, 2019). These investigations have found numerous factors likely to impact binding efficiency between virus and bacteria. Of note, Almand et al. (2017) demonstrated the impact of bacterial growth conditions on binding efficiency to HuNoV. Specifically, the authors observed a significant decrease in binding (i.e., less than 90% bacteria bound to HuNoV) when bacteria were grown in nutrient-dense media such as chopped meat or tryptic soy broth (TSB) with blood compared to minimal media or TSB. Almand et al. (2017) posit this difference in binding is due to differential expression of bacterial genes (e.g., possibly reduced expression of HBGA-like moieties on the cell membrane) as a result of media formulation. Standard culture media conditions were used for both B. cereus and E. cloacae; thus, it is unlikely to be a significant factor in the present study. Meanwhile, Li et al. (2015) also identified large variations in HBGA expression levels within a single bacterial strain between cultural batches, indicating the need for future studies in this area.

As indicated previously, studies have investigated how these interactions impact norovirus and norovirus surrogate response to heat stress. Li et al. (2015) published that when HuNoV-like particles (VLPs) were in association with HBGA-like expressing bacteria, the VLPs had higher antigen integrity than in association with bacteria without the antigen. The researchers indicate that this interaction could protect the virus from heat stress. Li et al. (2017) conducted a similar study and observed that when Tulane virus (TV) was bound to HBGA-like expressing bacteria, this interaction did not protect it from heat stress. The conflicting results in these two studies can be attributed to the different viruses used, human norovirus VLPs versus TV. The study that utilized VLPs had the limitation of a binding assay, which does not completely indicate infectivity whereas the TV study could measure virus infectivity via culture to demonstrate if the virus was denatured or not. The discrepancy between the studies testifies to how complex these interactions may be.

The results in this study indicate no consistently significant differences in viral inactivation on BB by aqueous ozone while in association with bacteria. A singular significant difference was observed for MNV in association with B. cereus after a 40 min ozone wash treatment. The difference in reduction could indicate that when MNV associates with Gram-positive bacteria, the virus is more susceptible to inactivation due to ozone, or that when in association with Gram-negative bacteria, it neither aids nor enhances the inactivation. The differences in the specific interaction that occurs between MNV and bacteria could be based on Gram type.

As indicated previously, MNV binds to compounds such as sialic acids as opposed to HBGA. While we did not define the specific interaction between MNV and E. cloacae or B. cereus, we can assume it likely differs from the interaction that HuNoV would have with these bacteria. Similar to the finding of HBGA-like substances on bacterial cell membranes, bacteria can also express some forms of sialic acids, including polysialic acids which can be found within the capsular polysaccharides of some pathogenic bacteria (Cress et al., 2014; Varki et al., 2015). These capsular polysaccharides are complex consisting of several carbohydrate moieties, and not all are particularly well-characterized (Cress et al., 2014). Recently, Zhu et al. (2020) identified a bacterium in sewage that expresses a ganglioside-linked terminal sialic acid (GD1a) specifically recognized by MNV. The authors reported a decrease in MNV infectivity when associated with the bacterium, presumably due to obstruction of MNV binding thus interfering with viral internalization; however, this was also dependent on the strain of MNV.

Interaction of MNV with other glycans expressed on bacteria could also prevent internalization and subsequent infection due to physical obstruction. There is even the potential for the interaction with a bacterium to result in destabilization of the MNV capsid or interference with the uncoating process. Overall, the aforementioned complexities related to virus-bacteria interactions indicate that further research is needed to elucidate the interactions which occurred in the present study.

Conclusion

An ozone wash does not significantly increase inactivation of viruses on fresh produce compared to the use of water alone under the conditions presented here. Additionally, the two different attachment times to produce had no significant difference in viral inactivation by ozone. Ozone does, however, have the potential to prevent cross-contamination. The surface of the produce could play a key role in the extent of inactivation of microbes and needs to be considered when evaluating sanitizers. Furthermore, the interaction between viruses and bacteria ge.nerally had no significant impact on viral inactivation when compared to the virus only results. There is a possibility that viruses in association with Gram-positive bacteria may be more susceptible to inactivation by ozone, but this is only speculation and further investigation is needed. Overall, variables impacting ozone wash effectiveness should be considered when implementing produce wash sanitation systems within food service operations.