Although wildlife intrusion and untreated manure have been associated with microbial contamination of produce, relatively few studies have examined the survival of Escherichia coli on produce under field conditions following contamination (e.g., via splash from wildlife feces). This experimental study was performed to estimate the die-off rate of E. coli on preharvest lettuce following contamination with a fecal slurry. During August 2015, field-grown lettuce was inoculated via pipette with a fecal slurry that was spiked with a three-strain cocktail of rifampin-resistant nonpathogenic E. coli. Ten lettuce heads were harvested at each of 13 time points following inoculation (0, 2.5, 5, and 24 h after inoculation and every 24 h thereafter until day 10). The most probable number (MPN) of E. coli on each lettuce head was determined, and die-off rates were estimated. The relationship between sample time and the log MPN of E. coli per head was modeled using a segmented linear model. This model had a breakpoint at 106 h (95% confidence interval = 69, 142 h) after inoculation, with a daily decrease of 0.70 and 0.19 log MPN for 0 to 106 h and 106 to 240 h following inoculation, respectively. These findings are consistent with die-off rates obtained in similar studies that assessed E. coli survival on produce following irrigation. Overall, these findings provide die-off rates for E. coli on lettuce that can be used in future quantitative risk assessments.

Between 2003 and 2012, Escherichia coli O157 outbreaks reported in the United States sickened 4,928, hospitalized 1,272, and killed 33 people (24). Recent E. coli O157 outbreaks have been associated with leafy greens (8, 15, 48), including a 2006 outbreak linked to fresh spinach that sickened 199, hospitalized 102, and killed 3 people throughout the United States (7). Microbial contamination of fresh produce, including leafy greens, can occur in the field (2, 16, 36), in processing environments (e.g., packing houses or fresh-cut operations) (16, 36), and immediately prior to consumption (e.g., in the home) (36). Multiple foodborne disease outbreaks associated with leafy greens also have been traced back to probable preharvest contamination events (1, 10, 18, 50). Thus, an understanding of the survival and transmission of foodborne pathogens in the preharvest environment is essential for developing effective and feasible strategies for reducing the foodborne disease risks associated with the consumption of produce.

Foodborne pathogens, including E. coli O157 and other Shiga toxin–producing E. coli strains, have been isolated from a variety of wild and domestic animals, indicating the potential of these animals to serve as a source of microbial contamination in the produce preharvest environment (5, 17, 29, 31, 32). Pathogens present in wildlife scat and untreated manure can be transferred to produce via defecation or application of manure to produce fields (2, 14, 19, 41). Atwill et al. (2) found that E. coli O157:H7 in simulated wildlife feces could be transferred to field-grown lettuce via splash during foliar irrigation. The use of fecally contaminated water for irrigation or frost protection also can be a direct route of produce contamination (20, 60). Cases of foodborne disease have been associated with wildlife intrusion into produce fields (28, 30, 33, 34) and the use of contaminated surface water for produce production (21, 40, 50). However, pathogen populations that transfer to produce die off over time under field conditions (6, 26, 42, 60). Thus, die-off rates can be used in quantitative risk assessments to identify potential intervention and control strategies for reducing food safety risks associated with fresh produce consumption. For example, die-off rates can be used in risk models to estimate levels of contamination on produce at specific times following potential contamination events (52).

Researchers have investigated bacterial die-off rates on field-grown produce and reported mean die-off rates for E. coli ranging from 0.4 to 1.64 log CFU day−1 (26, 42, 49, 60). Wood et al. (60) observed E. coli O157:H7 die-off rates of 0.54 to 1.64 log CFU day−1 on field-grown spinach in Nova Scotia, Canada. Daily die-off rates for Salmonella on field-grown spinach and lettuce in the United Kingdom ranged from 0.43 to 0.76 log CFU day−1 (26). Variations in bacterial die-off rates on produce have been associated with multiple factors, including plant health and leaf age (23, 55), environmental conditions (60, 61), and pathogen transfer matrix. Wood et al. (60) found that the time to reach the detection limit for E. coli O157:H7 on spinach grown in full sun and in partial shade was 72 to 100 hours and >150 hours, respectively. Because of the variability in previously reported die-off rates for E. coli on preharvest produce, more research is needed to evaluate existing data and generate new data that can be used to determine appropriate risk management strategies for reducing risks associated with the consumption of produce. The present study was conducted to generate experimental data on the die-off rate of E. coli on fresh produce under field conditions in the northeastern United States and to subsequently compare the observed die-off rate to previously reported rates for E. coli on preharvest produce.

Field set up

This field study was conducted in a romaine lettuce (Lactuca sativa L. var. longifolia cv. Green Towers; Harris Seeds, Rochester, NY) field at the Homer C. Thompson Vegetable Research Farm (Freeville, NY). Throughout the growing season a combination of tine weeding, hand weeding, and rototilling was used to thin the lettuce heads (at ca. 4 weeks) and weed the study field. The field consisted of a planted area (8.5 by 59.5 m) with five longitudinal beds (each 1.2 m wide) separated by 0.6-m furrows (see Fig. S1 in the Supplemental Material). Each bed consisted of four rows of seeds planted 0.4 m apart (20 rows total). Seeds were planted with a mechanical seeder (Monosem, Edwardsville, KS), and a 1.5-in. (3.81-cm) seeding rate was achieved. The field was surrounded by a bare ground buffer (at least 3.1 m) on each side. Overhead impact sprinklers were spaced around the field (with ca. 15 m between sprinklers); irrigation occurred as needed up to 1 week before inoculation of the lettuce with E. coli.

One hundred thirty lettuce heads growing in the study field were randomly selected for inclusion in the study and randomly divided into 1 of 13 treatment groups. Each treatment group of 10 lettuce heads was harvested and quantitatively tested for E. coli at a given time point after inoculation of the lettuce heads; the time points were 0, 2.5, 5, and 24 h after inoculation and every 24 h thereafter until day 10.

Bacterial strains

Three rifampin-resistant nonpathogenic E. coli strains (TVS 353, TVS 354, and TVS 355; University of California, Davis) (54) were used to prepare the three-strain cocktail for lettuce inoculation. Each strain was grown in duplicate on tryptic soy agar plates (TSA; BD, Franklin Lakes, NJ) at 37°C to stationary phase (18 to 24 h). Each plate was then flooded with 10 mL of phosphate-buffered saline (PBS), and the cells were resuspended using a 10-μL loop and 10-mL pipette (Stripette, Corning, Inc., Corning, NY). Bacterial suspensions were separately transferred into 15-mL conical centrifuge tubes (Falcon, Corning) and centrifuged at 2,500 ×g (for 5 min), and the culture supernatant was removed. The pellet was washed twice with 10 mL of PBS and then resuspended in 5 mL of PBS. This bacterial suspension was diluted 1:32 in PBS, and the optical density at 600 nm was measured. Based on this value, the culture was diluted in PBS to achieve ca. 1010 CFU mL−1.

To assess the potential for false-positive results due to naturally occurring rifampin-resistant E. coli, sampling was performed prior to the start of the study. Three composite soil samples, three vegetation samples, and four water samples were collected from nearby environments, including Fall Creek, the source of irrigation water used in this study. Soil and vegetation samples were diluted 1:2 in PBS and then serially diluted in duplicate to 10−11 in tryptic soy broth (BD) supplemented with 100 mg L−1 rifampin (EMD Chemicals, San Diego, CA) (TSB+R). After incubation for 18 to 24 h at 37°C, 3 μL of each dilution was cross-streaked on E. coli ChromAgar (DRG International, Springfield, NJ) supplemented with 100 mg L−1 rifampin (ECC+R). The ECC+R plates were then incubated at 42°C for 18 to 24 h. Water samples were processed as described by Weller et al. (59). A 250-mL water sample was passed through a 0.45-μm-pore-size filter unit (Nalgene, Rochester, NY). The filter was then transferred to a Whirl-Pak bag (Nasco, Fort Atkinson, WI) and enriched with 90 mL of TSB+R. Following incubation at 37°C for 18 to 24 h, 50 μL of the enrichment culture was streaked onto ECC+R plates, which were incubated at 42°C for 18 to 24 h. Although one soil sample was positive for rifampin-resistant E. coli after plating on ECC+R, we were not able to confirm this culture as E. coli by using the clpX PCR described below.

Although the use of nonpathogenic surrogate E. coli strains (rather than pathogenic wild-type strains) may be considered a drawback of this study, pathogenic E. coli could not be used because of biosecurity concerns. The three-strain E. coli cocktail was selected for use because others (22, 54) have found that this cocktail has greater environmental fitness than do individual attenuated E. coli O157:H7 strains. Thus, the die-off rates for the nonpathogenic E. coli strains in our study provide conservative estimates for pathogenic E. coli die-off on preharvest field-grown lettuce. Moreover, Gutiérrez-Rodríguez et al. (22) reported that the survival and persistence of pathogenic and nonpathogenic E. coli was strain dependent. Thus, the use of nonpathogenic surrogates can provide valuable information for further defining E. coli strain variability with regard to survival on produce.

Fecal slurry preparation

Laboratory rabbit feces (Oryctolagus cuniculus; CoVance, Princeton, NJ) were used as a proxy for wildlife feces in the present study. Other researchers (5, 17, 32) have identified wild and domestic rabbits as reservoirs for pathogenic E. coli. Fifty grams of feces, 200 mL of PBS, and 2.5 mL of the three-strain E. coli cocktail (0.833 mL of each strain culture) were combined in a sterile filter Whirl-Pak bag and hand massaged for 5 min. The solid matter was then separated from the liquid portion by pipetting the liquid portion (the fecal slurry) into a 50-mL Falcon tube, which was then stored overnight at 4°C. The final level of E. coli in the fecal slurry (3.5 × 108 CFU mL−1) was confirmed on the morning of inoculation by diluting, in triplicate, a 1-mL aliquot of the fecal slurry with PBS and spiral plating 50 μL of the 10−3, 10−4, and 10−5 dilutions on TSA plates supplemented with 100 mg L−1 rifampin. Plates were incubated at 37°C for 18 to 24 h, and colonies were enumerated using a Q-Count (Advanced Instruments, Norwood, MA).

Inoculation

Each lettuce head was inoculated by pipetting 1 mL of fecal slurry onto the northernmost lettuce leaf from a height of ca. 7 cm. Because of the volume of inoculum and the lettuce growth structure and leaf shape, the slurry tended to spread across the inoculated leaf toward and onto the stem and then drip onto the lower leaves. The inoculation method used in this study is not representative of all feces-related contamination events in terms of bacteria (rifampin-resistant nonpathogenic E. coli), source (rabbit feces), or deposition (in a single spot on the lettuce head as a slurry with a high bacterial level). However, our experimental design allowed us to control for confounding factors (e.g., feces source and location of inoculum on the lettuce leaf) and to track E. coli die-off in the preharvest environment. For example, although E. coli contamination of fresh produce is likely to occur at much lower levels than that used in this study, a higher level was used here to ensure comparability with previous studies (27, 42, 49). A high initial inoculum was also used to allow accurate quantification of die-off (49), which was expected to be >4 log over the 10-day time frame (6, 42, 60). In fact, in their review of studies that examined pathogen die-off on produce, Snellman et al. (49) included only those studies in which a high initial inoculum was used because of the difficulty in determining cell densities accurately at low inoculum levels.

Harvest

Inoculation occurred 84 days after seeding, and harvest occurred at predetermined time points following inoculation (0, 2.5, 5, and 24 h after inoculation and every 24 h thereafter until day 10). Lettuce heads were harvested by teams of two people, a bagger and a harvester, each of whom wore gloves. Lettuce heads were harvested with gloved hands using a food grade knife. Gloves were changed between collection of each lettuce head, and the knife was decontaminated with a 10% bleach wipe followed by a 70% ethanol wipe. A total of 130 lettuce heads were collected (10 heads per time point). All heads were placed in prelabeled Whirl-Pak bags, stored at 4°C, and processed within 3 h of harvest.

Enumeration of E. coli on lettuce

The enumeration methods used in this study were adapted from those of Atwill et al. (2). A 600-mL volume of PBS was added directly to each of the Whirl-Pak bags containing a lettuce head, and the bags were then hand massaged for 1 min. Rifampin-resistant E. coli was enumerated using a most-probable-number (MPN) method with six dilutions tested in duplicate. A 1-mL aliquot of sample suspension was transferred into each of 2 wells in a 12-well deep-well plate (VWR International, Radnor, PA); both wells also contained 9 mL of TSB+R. Five serial 100-fold dilutions (0.1 mL into 9.9 mL of TSB+R) were subsequently made starting from each of the two initial wells. Following incubation for 24 h at 37°C, 3 μL from each well was streaked onto ECC+R plates, which were then incubated at 42°C for 18 to 24 h. Blue colonies indicated the presumptive presence of one of the inoculation strains (TVS 353, TVS 354, or TVS 355). Detection of the inoculation strains was confirmed for 10% of presumptive-positive lettuce heads using PCR amplification and Sanger sequencing of the E. coli clpX gene as described in Walk et al. (58). Only a subset of isolates from positive lettuce heads was tested by clpX PCR and sequencing because all E. coli isolates tested were confirmed as a clpX allelic type that matched one of the inoculation strains. The MPN of cells per head was calculated as described by Cochran (9). The R script used to implement the method outlined by Cochran (9) is described in the Supplemental Material (Material S1).

Statistical analysis

All statistical analyses were performed in R (version 3.1, R Core Team, Vienna, Austria). The cutoff for significant results was P = 0.05 for all analyses. Die-off was visualized by plotting the mean level (log MPN per head) against time. Die-off per unit time between each time point (e.g., between 0 and 2.5 h, between 2.5 and 5 h) and each day (e.g., between 0 and 24 h, between 24 and 48 h) was calculated using the formula

\(\def\upalpha{\unicode[Times]{x3B1}}\)\(\def\upbeta{\unicode[Times]{x3B2}}\)\(\def\upgamma{\unicode[Times]{x3B3}}\)\(\def\updelta{\unicode[Times]{x3B4}}\)\(\def\upvarepsilon{\unicode[Times]{x3B5}}\)\(\def\upzeta{\unicode[Times]{x3B6}}\)\(\def\upeta{\unicode[Times]{x3B7}}\)\(\def\uptheta{\unicode[Times]{x3B8}}\)\(\def\upiota{\unicode[Times]{x3B9}}\)\(\def\upkappa{\unicode[Times]{x3BA}}\)\(\def\uplambda{\unicode[Times]{x3BB}}\)\(\def\upmu{\unicode[Times]{x3BC}}\)\(\def\upnu{\unicode[Times]{x3BD}}\)\(\def\upxi{\unicode[Times]{x3BE}}\)\(\def\upomicron{\unicode[Times]{x3BF}}\)\(\def\uppi{\unicode[Times]{x3C0}}\)\(\def\uprho{\unicode[Times]{x3C1}}\)\(\def\upsigma{\unicodeTimes]{x3C3}}\)\(\def\uptau{\unicode[Times]{x3C4}}\)\(\def\upupsilon{\unicode[Times]{x3C5}}\)\(\def\upphi{\unicode[Times]{x3C6}}\)\(\def\upchi{\unicode[Times]{x3C7}}\)\(\def\uppsy{\unicode[Times]{x3C8}}\)\(\def\upomega{\unicode[Times]{x3C9}}\)\(\def\bialpha{\boldsymbol{\alpha}}\)\(\def\bibeta{\boldsymbol{\beta}}\)\(\def\bigamma{\boldsymbol{\gamma}}\)\(\def\bidelta{\boldsymbol{\delta}}\)\(\def\bivarepsilon{\boldsymbol{\varepsilon}}\)\(\def\bizeta{\boldsymbol{\zeta}}\)\(\def\bieta{\boldsymbol{\eta}}\)\(\def\bitheta{\boldsymbol{\theta}}\)\(\def\biiota{\boldsymbol{\iota}}\)\(\def\bikappa{\boldsymbol{\kappa}}\)\(\def\bilambda{\boldsymbol{\lambda}}\)\(\def\bimu{\boldsymbol{\mu}}\)\(\def\binu{\boldsymbol{\nu}}\)\(\def\bixi{\boldsymbol{\xi}}\)\(\def\biomicron{\boldsymbol{\micron}}\)\(\def\bipi{\boldsymbol{\pi}}\)\(\def\birho{\boldsymbol{\rho}}\)\(\def\bisigma{\boldsymbol{\sigma}}\)\(\def\bitau{\boldsymbol{\tau}}\)\(\def\biupsilon{\boldsymbol{\upsilon}}\)\(\def\biphi{\boldsymbol{\phi}}\)\(\def\bichi{\boldsymbol{\chi}}\)\(\def\bipsy{\boldsymbol{\psy}}\)\(\def\biomega{\boldsymbol{\omega}}\)\begin{equation}{{\Delta {\rm{log\ MPN}}} \over {\Delta t}} = {\left( {{\rm{average\ log\ MPN}}} \right)_{t - 1}} - {\left( {{\rm{average\ log\ MPN}}} \right)_t}\end{equation}

where t − 1 and t are the two sampling points of interest and Δt is the length of time between these two sampling points. To statistically describe the change in log MPN per head over time, a linear regression model was developed. However, results of other studies indicate that microbial die-off can be biphasic (11, 39) and may be better modeled using a segmented linear model or a Weibull model. Using the segmented package in R (56, 57), we conducted a Davies test to determine whether the linear model included a nonconstant regression parameter (the breakpoint) and developed a segmented linear model. We then retested the segmented linear model to determine whether there was a second breakpoint. Using the nlsMicrobio package in R (4), we developed a Weibull model as parameterized by Mafart et al. (37). The formula for the Weibull model is nt = n0 − (t/δ)p, where nt is the log MPN of E. coli at time t, n0 is the log MPN of E. coli at time 0, δ is the time to the first decimal reduction, and p is a parameter that describes the concavity of the curve described by the model. Akaike's information criterion (AIC) was used to determine whether the linear, segmented linear, or Weibull model best fit the data.

Die-off was calculated using a log transformation (i.e., logarithm base 10) because this transformation is traditionally used by microbiologists, industry, and government stakeholders. To provide decay rates for modeling purposes (i.e., k in Ct = Ci + ekt, where Ct is level at time t and Ci is initial level), the slopes of the linear and segmented linear models are also reported using a natural logarithm transformation.

Weather conditions for the day of lettuce head inoculation and for 1 to 9 days after lettuce head inoculation are reported in the Supplemental Material (Table S1). Weather data were obtained from the Cornell University weather station (Rainwise, Trenton, NJ) located at the Homer C. Thompson Vegetable Research Farm as described by Weller et al. (59). Linear regression was used to statistically describe the relationship between the log MPN of E. coli per head and weather (this is referred to as the “weather model”). The dependent variable of the model was the log MPN of E. coli per head. The explanatory variables were (i) the period of time between inoculation and harvest (hours), (ii) mean temperature, relative humidity, and wind speed for the 24 h preceding harvest, (iii) total leaf wetness for the 24 h preceding harvest, and (iv) whether the lettuce head was harvested before or after the rain event (ca. 7.1 mm) that occurred 64 to 69 h after inoculation (post-rain). The interaction between hours and the post-rain variable was also included in the model. The full model was reduced by backward stepwise regression based on the AIC. Each variable was removed from the full model, and the AIC was determined. The variable whose removal resulted in the largest decrease in the AIC was removed from the model. This process was repeated until the removal of additional variables failed to reduce the AIC.

Observed die-off rate for E. coli was 0.52 log MPN day−1

During the 240 h between inoculation and harvest on day 10, the mean E. coli level decreased from 8.86 to 3.64 log MPN per head (Fig. 1), a 5.22-log reduction. On average, we observed a die-off rate of 0.52 log MPN per head day−1 (95% confidence interval [CI] = 0.17, 0.87; Table 1); this value falls within the range of previously reported daily die-off rates for E. coli on produce (0.4 to 1.64 log MPN day−1) (26, 42, 49, 60). The observed die-off rate was also similar to die-off rates that can be calculated using the findings reported by Barker-Reid et al. (3) for nonpathogenic E. coli on uninjured lettuce (0.44 log day−1) and Bezanson et al. (6) for E. coli O157:H7 on lettuce (0.56 log CFU day−1). Further comparisons of the die-off rate observed in this study and in previous studies is presented below.

FIGURE 1.

E. coli levels (log MPN per lettuce head) for each time point (e.g., 0, 2.5, and 5 h) reported as mean (gray points) and standard deviation (gray bars) and minimum and maximum (blue shading). The linear regression (A), segmented linear (B), and Weibull (C) models describe E. coli die-off over time.

FIGURE 1.

E. coli levels (log MPN per lettuce head) for each time point (e.g., 0, 2.5, and 5 h) reported as mean (gray points) and standard deviation (gray bars) and minimum and maximum (blue shading). The linear regression (A), segmented linear (B), and Weibull (C) models describe E. coli die-off over time.

Close modal
TABLE 1.

Average die-off of inoculated E. coli on lettuce heads grown under field conditions

Average die-off of inoculated E. coli on lettuce heads grown under field conditions
Average die-off of inoculated E. coli on lettuce heads grown under field conditions

E. coli is still detectable on the lettuce 10 days after inoculation

Ten days after inoculation, E. coli was still detectable on the lettuce heads (Fig. 1; mean ± SD = 3.64 ± 0.75 log MPN). By comparison, previous studies with high inoculum levels for E. coli (105 to 109 CFU mL−1) reported a range of times until the E. coli levels dropped below the detection limits for the respective methods used (12, 13, 38, 42). For example, in a study in Georgia, United States (27), E. coli O157:H7 was still detectable on lettuce 77 days after the field was fertilized with contaminated manure (inoculum of 107 CFU g−1) and 77 days after irrigation with contaminated water (inoculum of 105 CFU mL−1). However, in another study in Georgia (12), E. coli was not detectable by enrichment 7 days after irrigation with contaminated water (inoculum of 106 CFU mL−1). In comparison, 82% of lettuce samples in a multiyear study in California (42) had less than 10 E. coli O157:H7 cells per head 7 days after inoculation with 107 CFU mL−1. Other researchers have also found that environmental conditions, including weather and season (53, 54, 60, 61), appear to be associated with E. coli survival on preharvest produce. Xu et al. (61) found that E. coli populations on field-grown spinach increased by up to 1 log unit after <20 mm of rain but decreased after >35 mm of rain. Studies with multiple replicates over time (e.g., multiple growing seasons, multiple years, or staggered planting of fields) and/or space (e.g., trials on different farms and in different regions) will be needed to further assess and quantify the effect of environmental conditions on E. coli die-off on field-grown produce.

Segmented linear model indicates that E. coli die-off follows a biphasic pattern with rapid initial die-off over the first approximately 100 h and more gradual die-off thereafter

A linear regression model built with the data generated here predicted a mean daily E. coli decrease of 0.46 log MPN per lettuce head (95% CI = 0.38, 0.50 log MPN; Table 2 and Fig. 1A), which is similar to the daily die-off rate reported previously (26, 42, 49, 60). The linear model accounted for 66% of all variation in the observed log MPN decrease per head (R2 = 0.66). However, the raw data suggested a biphasic decrease (Fig. 1B), which may be better represented by a segmented linear model. Using a Davies test, we determined that there was a nonconstant regression parameter in the linear predictor (P < 0.001). Therefore, we developed a segmented linear model (Table 2 and Fig. 1B) with a breakpoint at 106 h (95% CI = 69, 142 h). After visually examining the data (Fig. 1B), we thought there might be a second breakpoint in the first 48 h immediately following inoculation. Therefore, we ran a second Davies test using the segmented linear model and identified a second breakpoint at ca. 5 h. However, this breakpoint was not significant (P = 0.38) and was not included in the final segmented model (Table 2). The segmented linear model predicts mean daily E. coli decreases of 0.70 log MPN (95% CI = 0.55, 0.86 log MPN) and 0.19 log MPN (95% CI = 0.05, 0.36 log MPN) for 0 to 106 h and 106 to 240 h, respectively (Table 2). The segmented linear model accounted for 71% of all variation in the decrease in log MPN per head observed in this study (R2 = 0.71). Because others have found that Weibull models accurately describe bacterial die-off (37, 39), we also developed a Weibull model. The formula for the Weibull model is nt = n0 − (t/δ)p, where nt is the log MPN of E. coli at time t, n0 is the log MPN of E. coli at time 0, δ is the time to the first decimal reduction, and p describes the concavity of the curve described by the model. The Weibull model for the study reported here is nt = 8.62 − (t/10.21)0.50 (Table 3). The AIC for the segmented linear and Weibull models was the same (380) and was lower than the AIC for the linear model (396; Tables 2 and 3). This suggests that the segmented model and Weibull model are comparable and better fit the data than the linear model. However, because the parameters of the segmented model have a more intuitive interpretation than do the parameters of the Weibull model, we focus on the segmented model in our discussion.

TABLE 2.

Parameters for two models that characterize the relationship between hours from inoculation to harvest and the E. coli level per lettuce heada

Parameters for two models that characterize the relationship between hours from inoculation to harvest and the E. coli level per lettuce heada
Parameters for two models that characterize the relationship between hours from inoculation to harvest and the E. coli level per lettuce heada
TABLE 3.

Parameters for a Weibull model that statistically characterizes the relationship between hours from inoculation to harvest and the E. coli level per lettuce head

Parameters for a Weibull model that statistically characterizes the relationship between hours from inoculation to harvest and the E. coli level per lettuce head
Parameters for a Weibull model that statistically characterizes the relationship between hours from inoculation to harvest and the E. coli level per lettuce head

As part of our analyses, we also examined the relationship between the E. coli log MPN per head and weather using a linear regression model. Of the seven factors that were included in the full model, five factors were retained in the final model (Table 4). The weather model accounted for 70% of all variation in the decrease in log MPN per head observed in this study (R2 = 0.70). The AIC for the weather model is 381, and the Bayesian information criterion for the weather model is 401.

TABLE 4.

Parameters for a linear regression model that characterizes the relationship between hours from inoculation to harvest, weather, and the E. coli level per lettuce heada

Parameters for a linear regression model that characterizes the relationship between hours from inoculation to harvest, weather, and the E. coli level per lettuce heada
Parameters for a linear regression model that characterizes the relationship between hours from inoculation to harvest, weather, and the E. coli level per lettuce heada

Our findings suggest that during the first ca. 100 h after inoculation there is a period of rapid E. coli die-off, which has also been observed in previous studies (23, 42, 54, 60). Our findings that the die-off was biphasic and was best represented by the segmented linear and Weibull models are also consistent with previous studies on E. coli die-off conducted in agricultural (6, 39, 47) and nonagricultural (11, 25) environments. McKellar et al. (39) evaluated various approaches for modeling E. coli die-off on field-grown lettuce using previously published data sets and found that E. coli die-off followed a biphasic pattern with a rapid initial decline. Because we collected data for only 10 days and included only four data points for the first 48 h immediately after inoculation, we were not able to model die-off after 10 days or determine whether additional breakpoints occurred during the first 48 h immediately after inoculation. Future studies should therefore include (i) samples collected for more than 10 days and (ii) additional data points obtained during the first 48 h after inoculation. However, our data indicate that the time immediately after inoculation (the first ca. 100 h) is the most important for E. coli reduction due to rapid die-off during this time.

Various mechanisms may explain the biphasic die-off pattern observed here and in other studies (35, 39, 42, 46). One possible explanation is heterogeneity within the microbial population in the inoculum (e.g., use of multiple strains or heterogeneous bacterial populations in stationary phase) and adaptation of the surviving microbial population to field conditions. Variation in environmental conditions (e.g., inner versus outer leaves) could also cause the biphasic pattern observed in this study. Peleg (45) postulated that microbial die-off is driven by environmental conditions, and as a result, exposed populations (e.g., on outer leaves) decline more rapidly than protected populations (e.g., on inner leaves). In previous research, contamination of inner younger leaves and other protected areas (e.g., shaded leaves) facilitated survival (12, 44, 55, 60). Other researchers have found associations between environmental conditions, such as UV radiation (43, 51, 60) and moisture levels (3, 44), and microbial die-off rates. Although analysis of weather patterns in the present study revealed no significant association between temperature and die-off, there appeared to be a significant association between precipitation and die-off (Table 4). The breakpoint identified in the segmented model (at ca. 106 h, 95% CI = 69, 142 h) occurred shortly after a moderate rain event (ca. 7.1 mm at 64 to 69 h). According to linear regression analysis the die-off rate was significantly lower for lettuce heads harvested after the rain event, which occurred between 64 and 69 h after inoculation, than for heads harvested before the rain event (Table 4). Although the moderate rain event may have washed bacteria off of the leaves, other factors including relative humidity and leaf wetness also were associated with bacterial die-off (Table 4). Because our study was conducted over the course of a single growing season (and we thus lacked a comparison group), the impact of weather is difficult to separate from the impact of time after inoculation. Because the weather model reported in Table 4 accounts for slightly less variation in the data (R2 = 70%) than does the segmented model (R2 = 71%), the observed biphasic pattern in microbial die-off could be explained almost equally well with or without the explicit consideration of weather. A biphasic die-off pattern for E. coli on produce has been reported previously (39) based on experiments conducted under presumably different environmental conditions including weather. Further studies with larger data sets collected over multiple growing seasons are needed to confirm our findings and build upon the data presented here.

Die-off rates reported here and in other studies appear to be comparable

The die-off rates that were observed (0.52 log MPN day−1) and calculated (0.70 and 0.19 log MPN day−1 for 0 to 106 h and 106 to 240 h, respectively) as part of this study are at the lower end of the range reported previously (0.4 to 1.64 log MPN day−1) (26, 42, 49, 60). However, in other studies (3, 6, 26, 60) die-off rates of 0.70 log MPN day−1 and lower were found. McKellar et al. (39) found that die-off rates were positively associated with inoculum level. Because the inoculum levels for natural contamination events are likely lower than the inoculum levels used in this and other studies (3, 26, 42, 60), die-off rates following actual contamination events may be lower than those reported previously. Thus, although die-off rates similar to those reported here provide conservative estimates for calculating time-to-harvest intervals, their use may overestimate die-off following contamination with low levels of E. coli or other similar bacteria. However, the daily die-off rates reported here and in previous studies (26, 42, 49, 60) were all within an approximately 1-log range, even though the studies had different designs (e.g., use of pathogenic and nonpathogenic E. coli strains, produce type, and inoculation procedures) and were performed under different conditions (e.g., weather and soil type). This similarity in rates suggests that the die-off rates reported to date are reasonable and comparable and can be used in quantitative risk assessments to evaluate the public health impact of preharvest risk management strategies.

Overall, the findings reported here are consistent with the die-off rates reported previously. Thus, the die-off rates reported here and in similar studies can be used in quantitative risk assessments and may therefore contribute to the development of effective risk management strategies, including time-to-harvest recommendations after potential contamination events. The present study is also the first to calculate die-off rates for field-grown preharvest produce in New York State and thus provides a foundational data set on which future studies can build.

This work was supported by the U.S. Food and Drug Administration through a contract with the Research Triangle Institute. We are grateful for the technical assistance of Maureen Gunderson, Rick Randolph, and Steve McKay, the statistical support of Dr. Erika Mudrak, and the technical review of Dr. Jane M. Van Doren. We are also grateful to Jeffrey Black, Miquela Hanselman, Thomas Hilchey, Alexander Alles, Courtenay Simmons, Aljosa Trmcic, Kanika Chauhan, Rachel Miller, Rachel Evanowski, Laura Carroll, Veronica Guariglia, Thomas Denes, Lory Henderson, Ariel Buehler, Sarah Beno, Barbara Bowen, and Samuel Reichler for help in the field and laboratory. We also thank Dr. Trevor Suslow, Janneth Pinzon, and Chelsea Kaminski for providing the strains used in this study.

Supplemental material associated with this article can be found online at: https://doi.org/10.4315/0362.028X.JFP-16-419.s1.

1
Ackers
,
M.
,
B.
Mahon
,
E.
Leahy
,
B.
Goode
,
T.
Damrow
,
P.
Hayes
,
W.
Bibb
,
D.
Rice
,
T.
Barrett
,
L.
Hutwagner
,
P.
Griffin
, and
L.
Slutsker
.
1998
.
An outbreak of Escherichia coli O157:H7 infections associated with leaf lettuce consumption
.
J. Infect. Dis
.
177
:
1588
1593
.
2
Atwill
,
E. R.
,
J.
Chase
,
D.
Oryang
,
R. F.
Bond
,
S. T.
Koike
,
M. D.
Cahn
,
M.
Anderson
,
A.
Mokhtari
, and
S.
Dennis
.
2015
.
Transfer of Escherichia coli O157:H7 from simulated wildlife scat onto romaine lettuce during foliar irrigation
.
J. Food Prot
.
78
:
240
247
.
3
Barker-Reid
,
F.
,
D.
Harapas
,
S.
Engleitner
,
S.
Kreidl
,
R.
Holmes
, and
R.
Faggian
.
2009
.
Persistence of Escherichia coli on injured iceberg lettuce in the field, overhead irrigated with contaminated water
.
J. Food Prot
.
72
:
458
464
.
4
Baty
,
F.
, and
M. L.
Delignette-Muller
.
2013
.
nlsMicrobio: data sets and nonlinear regression models dedicated to predictive microbiology. R package version 0.0-1
. .
5
Bergholz
,
P. W.
,
J. D.
Noar
, and
D. H.
Buckley
.
2011
.
Environmental patterns are imposed on the population structure of Escherichia coli after fecal deposition
.
Appl. Environ. Microbiol
.
77
:
211
219
.
6
Bezanson
,
G.
,
P.
Delaquis
,
S.
Bach
,
R.
McKellar
,
E.
Topp
,
A.
Gill
,
B.
Blais
, and
M.
Gilmour
.
2012
.
Comparative examination of Escherichia coli O157:H7 survival on romaine lettuce and in soil at two independent experimental sites
.
J. Food Prot
.
75
:
480
487
.
7
Centers for Disease Control and Prevention
.
2006
.
Multistate outbreak of E. coli O157:H7 infections linked to fresh spinach (final update)
.
Available at: http://www.cdc.gov/ecoli/2006/spinach-10-2006.html. Accessed 3 April 2017
.
8
Centers for Disease Control and Prevention
.
2012
.
Multistate outbreak of E. coli O157:H7 infections linked to romaine lettuce (final update)
. .
9
Cochran
,
W.
1950
.
Estimation of bacterial densities by means of the “most probable number.”
Biometrics
6
:
105
116
.
10
Crawford
,
W.
,
M.
Baloch
, and
K.
Gerrity
.
2010
.
Environmental assessment: non-O157 Shiga toxin–producing E. coli (STEC). U.S
.
Food and Drug Administration
,
Washington, DC
. .
11
Easton
,
J. H.
,
J. J.
Gauthier
,
M. M.
Lalor
, and
R. E.
Pitt
.
2005
.
Die-off of pathogenic E. coli O157:H7 in sewage contaminated waters
.
J. Am. Water Resour. Assoc
.
41
:
1187
1193
.
12
Erickson
,
M. C.
,
C. C.
Webb
,
J. C.
Diaz-Perez
,
S. C.
Phatak
,
J. J.
Silvoy
,
L.
Davey
,
A. S.
Payton
,
J.
Liao
,
L.
Ma
, and
M. P.
Doyle
.
2010
.
Surface and internalized Escherichia coli O157:H7 on field-grown spinach and lettuce treated with spray-contaminated irrigation water
.
J. Food Prot
.
73
:
1023
1029
.
13
Fonseca
,
J. M.
,
S. D.
Fallon
,
C.
Sanchez
, and
K. D.
Nolte
.
2011
.
Escherichia coli survival in lettuce fields following its introduction through different irrigation systems
.
J. Appl. Microbiol
.
110
:
893
902
.
14
Franz
,
E.
,
A. V.
Semenov
, and
A. H. C.
van Bruggen
.
2008
.
Modelling the contamination of lettuce with Escherichia coli O157:H7 from manure-amended soil and the effect of intervention strategies
.
J. Appl. Microbiol
.
105
:
1569
1584
.
15
Friesema
,
I.
,
G.
Sigmundsdottir
,
K.
van der Zwaluw
,
A.
Heuvelink
,
B.
Schimmer
,
C.
de Jager
,
B.
Rump
,
H.
Briem
,
H.
Hardardottir
,
A.
Atladottir
,
E.
Gudmundsdottir
, and
W.
van Pelt
.
2008
.
An international outbreak of Shiga toxin–producing Escherichia coli O157 infection due to lettuce, September–October 2007
.
Eur. Commun. Dis. Bull
.
13
:
3029
3035
.
16
Gagliardi
,
J. V.
,
P. D.
Millner
,
G.
Lester
, and
D.
Ingram
.
2003
.
On-farm and postharvest processing sources of bacterial contamination to melon rinds
.
J. Food Prot
.
66
:
82
87
.
17
García
,
A.
, and
J. G.
Fox
.
2003
.
The rabbit as a new reservoir host of enterohemorrhagic Escherichia coli
.
Emerg. Infect. Dis
.
9
:
1592
1597
.
18
Gelting
,
R. J.
,
M.
Baloch
,
M.
Zarate-Bermudez
, and
C.
Selman
.
2011
.
Irrigation water issues potentially related to the 2006 multistate E. coli O157:H7 outbreak associated with spinach
.
Agric. Water Manag
.
98
:
1395
1402
.
19
Généreux
,
M.
,
M. J.
Breton
,
J. M.
Fairbrother
,
P.
Fravalo
, and
C.
Côté
.
2015
.
Persistence of indicator and pathogenic microorganisms in broccoli following manure spreading and irrigation with fecally contaminated water: field experiment
.
J. Food Prot
.
78
:
1776
1784
.
20
Gorman
,
S.
2014
.
Transfer and survival of microorganisms to produce from surface irrigation water
.
M.S. thesis
.
University of Tennessee
,
Knoxville
.
Available at: http://trace.tennessee.edu/utk_gradthes/2819/. Accessed 3 April 2017
.
21
Greene
,
S. K.
,
E. R.
Daly
,
E. A.
Talbot
,
L. J.
Demma
,
S.
Holzbauer
,
N. J.
Patel
,
T. A.
Hill
,
M. O.
Walderhaug
,
R. M.
Hoekstra
,
M. F.
Lynch
, and
J. A.
Painter
.
2008
.
Recurrent multistate outbreak of Salmonella Newport associated with tomatoes from contaminated fields, 2005
.
Epidemiol. Infect
.
136
:
157
165
.
22
Gutiérrez-Rodríguez
,
E.
,
A.
Gundersen
,
A. O.
Sbodio
, and
T. V.
Suslow
.
2012
.
Variable agronomic practices, cultivar, strain source and initial contamination dose differentially affect survival of on spinach
.
J. Appl. Microbiol
.
112
:
109
118
.
23
Harapas
,
D.
,
R.
Premier
,
B.
Tomkins
,
P.
Franz
, and
S.
Ajlouni
.
2010
.
Persistence of Escherichia coli on injured vegetable plants
.
Int. J. Food Microbiol
.
138
:
232
237
.
24
Heiman
,
K. E.
,
R. K.
Mody
,
S. D.
Johnson
,
P. M.
Griffin
, and
L. H.
Gould
.
2015
.
Escherichia coli O157 outbreaks in the United States, 2003–2012
.
Emerg. Infect. Dis
.
21
:
1293
1301
.
25
Hellweger
,
F. L.
,
V.
Bucci
,
M. R.
Litman
,
A. Z.
Gu
, and
A.
Onnis-Hayden
.
2009
.
Biphasic decay kinetics of fecal bacteria in surface water not a density effect
.
J. Environ. Eng
.
135
:
372
376
.
26
Hutchison
,
M. L.
,
S. M.
Avery
, and
J. M.
Monaghan
.
2008
.
The air-borne distribution of zoonotic agents from livestock waste spreading and microbiological risk to fresh produce from contaminated irrigation sources
.
J. Appl. Microbiol
.
105
:
848
857
.
27
Islam
,
M.
,
M. P.
Doyle
,
S. C.
Phatak
,
P.
Millner
, and
X.
Jiang
.
2005
.
Persistence of enterohemorrhagic Escherichia coli O157:H7 in soil and on leaf lettuce and parsley grown in fields treated with contaminated manure composts or irrigation water
.
J. Food Prot
.
67
:
1365
1370
.
28
Jay
,
M. T.
,
M.
Cooley
,
D.
Carychao
,
G. W.
Wiscomb
,
R.
Sweitzer
,
L.
Crawford-Miksza
,
J.
Farrar
,
D. K.
Lau
,
J.
O'Connell
,
A.
Millington
,
R. V.
Asmundson
,
E. R.
Atwill
, and
R. E.
Mandrell
.
2007
.
Escherichia coli O157:H7 in feral swine near spinach fields and cattle, central California coast
.
Emerg. Infect. Dis
.
13
:
1908
1911
.
29
Jay-Russell
,
M. T.
,
A. F.
Hake
,
Y.
Bengson
,
A.
Thiptara
, and
T.
Nguyen
.
2014
.
Prevalence and characterization of Escherichia coli and Salmonella strains isolated from stray dog and coyote feces in a major leafy greens production region at the United States–Mexico border
.
PLoS ONE
9
(
11
):
e113433
. .
30
Kangas
,
S.
,
J.
Takkinen
,
M.
Hakkinen
,
U. M.
Nakari
,
T.
Johansson
,
H.
Henttonen
,
L.
Virtaluoto
,
A.
Siitonen
,
J.
Ollgren
, and
M.
Kuusi
.
2008
.
Yersinia pseudotuberculosis O:1 traced to raw carrots, Finland
.
Emerg. Infect. Dis
.
14
:
1959
1961
.
31
Kilonzo
,
C.
,
X.
Li
,
E. J.
Vivas
,
M. T.
Jay-Russell
,
K. L.
Fernandez
, and
E. R.
Atwill
.
2013
.
Fecal shedding of zoonotic food-borne pathogens by wild rodents in a major agricultural region of the central California coast
.
Appl. Environ. Microbiol
.
79
:
6337
6344
.
32
Kohler
,
R.
,
G.
Krause
,
L.
Beutin
,
R.
Stephan
, and
C.
Zweifel
.
2008
.
Shedding of food-borne pathogens and microbiological carcass contamination in rabbits at slaughter
.
Vet. Microbiol
.
132
:
149
157
.
33
Kwan
,
P. S. L.
,
C.
Xavier
,
M.
Santovenia
,
J.
Pruckler
,
S.
Stroika
,
K.
Joyce
,
T.
Gardner
,
P. I.
Fields
,
J.
McLaughlin
,
R. V.
Tauxe
, and
C.
Fitzgerald
.
2014
.
Multilocus sequence typing confirms wild birds as the source of a Campylobacter outbreak associated with the consumption of raw peas
.
Appl. Environ. Microbiol
.
80
:
4540
4546
.
34
Laidler
,
M. R.
,
M.
Tourdjman
,
G. L.
Buser
,
T.
Hostetler
,
K. K.
Repp
,
R.
Leman
,
M.
Samadpour
, and
W. E.
Keene
.
2013
.
Escherichia coli O157:H7 infections associated with consumption of locally grown strawberries contaminated by deer
.
Clin. Infect. Dis
.
57
:
1129
1134
.
35
Lundquist
,
E. J.
,
K. M.
Scow
,
L. E.
Jackson
,
S. L.
Uesugi
, and
C. R.
Johnson
.
1999
.
Rapid response of soil microbial communities from conventional, low input, and organic farming systems to a wet/dry cycle
.
Soil Biol. Biochem
.
31
:
1661
1675
.
36
Lynch
,
M. F.
,
R. V.
Tauxe
, and
C. W.
Hedberg
.
2009
.
The growing burden of foodborne outbreaks due to contaminated fresh produce: risks and opportunities
.
Epidemiol. Infect
.
137
:
307
315
.
37
Mafart
,
P.
,
O.
Couvert
,
S.
Gaillard
, and
I.
Leguerinel
.
2002
.
On calculating sterility in thermal preservation methods: application of the Weibull frequency distribution model
.
Int. J. Food Microbiol
.
72
:
107
113
.
38
Markland
,
S. M.
,
K. L.
Shortlidge
,
D. G.
Hoover
,
S.
Yaron
,
J.
Patel
,
A.
Singh
,
M.
Sharma
, and
K. E.
Kniel
.
2013
.
Survival of pathogenic Escherichia coli on basil, lettuce, and spinach
.
Zoonoses Public Health
60
:
563
571
.
39
McKellar
,
R. C.
,
F.
Peréz-Rodríguez
,
L. J.
Harris
,
A.
Moyne
,
B.
Blais
,
E.
Topp
,
G.
Bezanson
,
S.
Bach
, and
P.
Delaquis
.
2014
.
Evaluation of different approaches for modeling Escherichia coli O157:H7 survival on field lettuce
.
Int. J. Food Microbiol
.
184
:
74
85
.
40
Mody
,
R. K.
,
S.
Greene
,
L.
Gaul
,
A.
Sever
,
S.
Pichette
,
I.
Zambrana
,
T.
Dang
,
A.
Gass
,
R.
Wood
,
K.
Herman
,
L. B.
Cantwell
,
G.
Falkenhorst
,
K.
Wannemuehler
,
R. M.
Hoekstra
,
I.
McCullum
,
A.
Cone
,
L.
Franklin
,
J.
Austin
,
K.
Delea
,
C. B.
Behravesh
,
S. V.
Sodha
,
J. C.
Yee
,
B.
Emanuel
,
S. F.
Al-Khaldi
,
V.
Jefferson
,
I. T.
Williams
,
P. M.
Griffin
, and
D. L.
Swerdlow
.
2011
.
National outbreak of Salmonella serotype Saintpaul infections: importance of Texas restaurant investigations in implicating jalapeño peppers
.
PLoS ONE
6
(
2
):
e16579
. .
41
Monaghan
,
J. M. M.
, and
M. L. L.
Hutchison
.
2012
.
Distribution and decline of human pathogenic bacteria in soil after application in irrigation water and the potential for soil-splash–mediated dispersal onto fresh produce
.
J. Appl. Microbiol
.
112
:
1007
1019
.
42
Moyne
,
A.
,
M. R.
Sudarshana
,
T.
Blessington
,
S. T.
Koike
,
M. D.
Cahn
, and
L. J.
Harris
.
2011
.
Fate of Escherichia coli O157:H7 in field-inoculated lettuce
.
Food Microbiol
.
28
:
1417
1425
.
43
Oladeinde
,
A.
,
T.
Bohrmann
,
K.
Wong
,
S. T.
Purucker
,
K.
Bradshaw
,
R.
Brown
,
B.
Snyder
, and
M.
Molina
.
2014
.
Decay of fecal indicator bacterial populations and bovine-associated source-tracking markers in freshly deposited cow pats
.
Appl. Environ. Microbiol
.
80
:
110
118
.
44
Oliveira
,
M.
,
I.
Viñas
,
J.
Usall
,
M.
Anguera
, and
M.
Abadias
.
2012
.
Presence and survival of Escherichia coli O157:H7 on lettuce leaves and in soil treated with contaminated compost and irrigation water
.
Int. J. Food Microbiol
.
156
:
133
140
.
45
Peleg
,
M.
2006
.
Isothermal microbial heat inactivation
,
p
.
1
48
.
In
Advanced quantitative microbiology for foods and biosystems
.
CRC Press
,
Boca Raton, FL
.
46
Rogers
,
S. W.
,
M.
Donnelly
,
L.
Peed
,
C.
Kelty
,
S.
Mondal
,
Z.
Zhong
, and
O. C.
Shanks
.
2011
.
Decay of bacterial pathogens, fecal indicators, and real-time quantitative PCR genetic markers in manure-amended soils
.
Appl. Environ. Microbiol
.
77
:
4839
4848
.
47
Seidu
,
R.
,
I.
Sjølander
,
A.
Abubakari
,
D.
Amoah
,
J.
Larbi
, and
T.
Stenström
.
2013
.
Modeling the die-off of E. coli and Ascaris in wastewater-irrigated vegetables: implications for microbial health risk reduction associated with irrigation cessation
.
Water Sci. Technol
.
68
:
1013
1021
.
48
Slayton
,
R. B.
,
G.
Turabelidze
,
S. D.
Bennett
,
C.
Schwensohn
,
A. Q.
Yaffee
,
F.
Khan
,
C.
Butler
,
E.
Trees
,
T. L.
Ayers
,
M. L.
Davis
,
A. S.
Laufer
,
S.
Gladbach
,
I.
Williams
, and
L. B.
Gieraltowski
.
2013
.
Outbreak of Shiga toxin–producing Escherichia coli (STEC) O157:H7 associated with romaine lettuce consumption
,
2011
.
PLoS ONE
8
(
2
):
e55300
.
Available at: http://dx.plos.org/10.1371/journal.pone.0055300. Accessed 3 April 2017
.
49
Snellman
,
E.
,
M.
Fatica
,
K.
Ravaliya
, and
S.
Assar
.
2014
.
Memorandum to the file—review of microbial decay constants reported in field trials of contaminated produce, re standards for the growing, harvesting, packing, and holding of produce for human consumption
.
Federal document FDA-2011-N-0921-0992
. .
50
Söderström
,
A.
,
P.
Osterberg
,
A.
Lindqvist
,
B.
Jönsson
,
A.
Lindberg
,
S.
Blide Ulander
,
C.
Welinder-Olsson
,
S.
Löfdahl
,
B.
Kaijser
,
B.
De Jong
,
S.
Kühlmann-Berenzon
,
S.
Boqvist
,
E.
Eriksson
,
E.
Szanto
,
S.
Andersson
,
G.
Allestam
,
I.
Hedenström
,
L.
Ledet Muller
, and
Y.
Andersson
.
2008
.
A large Escherichia coli O157 outbreak in Sweden associated with locally produced lettuce. Foodborne Pathog
.
Dis
.
5
:
339
349
.
51
Steele
,
M.
, and
J.
Odumeru
.
2004
.
Irrigation water as source of foodborne pathogens on fruit and vegetables
.
J. Food Prot
.
67
:
2839
2849
.
52
Stine
,
S. W.
,
I.
Song
,
C. Y.
Choi
, and
C. P.
Gerba
.
2005
.
Application of microbial risk assessment to the development of standards for enteric pathogens in water used to irrigate fresh produce
.
J. Food Prot
.
68
:
913
918
.
53
Stine
,
S. W.
,
I.
Song
,
C. Y.
Choi
, and
C. P.
Gerba
.
2005
.
Effect of relative humidity on preharvest survival of bacterial and viral pathogens on the surface of cantaloupe, lettuce, and bell peppers
.
J. Food Prot
.
68
:
1352
1358
.
54
Tomás-Callejas
,
A.
,
G.
López-Velasco
,
A. B.
Camacho
,
F.
Artés
,
F.
Artés-Hernández
, and
T. V.
Suslow
.
2011
.
Survival and distribution of Escherichia coli on diverse fresh-cut baby leafy greens under preharvest through postharvest conditions
.
Int. J. Food Microbiol
.
151
:
216
222
.
55
Van der Linden
,
I.
,
B.
Cottyn
,
M.
Uyttendaele
,
G.
Vlaemynck
,
M.
Heyndrickx
, and
M.
Maes
.
2013
.
Survival of enteric pathogens during butterhead lettuce growth: crop stage, leaf age, and irrigation. Foodborne Pathog
.
Dis
.
10
:
485
491
.
56
Vito
,
M.
, and
R.
Muggeo
.
2003
.
Estimating regression models with unknown break-points
.
Stat. Med
.
22
:
3055
3071
.
57
Vito
,
M.
, and
R.
Muggeo
.
2008
.
Segmented: an R package to fit regression models with broken-line relationships
.
R News
8
:
20
25
.
58
Walk
,
S. T.
,
E. W.
Alm
,
D. M.
Gordon
,
J. L.
Ram
,
G.
Toranzos
,
J. M.
Tiedje
, and
T. S.
Whittam
.
2009
.
Cryptic lineages of the genus Escherichia
.
Appl. Environ. Microbiol
.
75
:
6534
6544
.
59
Weller
,
D.
,
M.
Wiedmann
, and
L. K.
Strawn
.
2015
.
Spatial and temporal factors associated with an increased prevalence of L. monocytogenes in spinach fields in New York State
.
Appl. Environ. Microbiol
.
81
:
6059
6069
.
60
Wood
,
J. D.
,
G. S.
Bezanson
,
R. J.
Gordon
, and
R.
Jamieson
.
2010
.
Population dynamics of Escherichia coli inoculated by irrigation into the phyllosphere of spinach grown under commercial production conditions
.
Int. J. Food Microbiol
.
143
:
198
204
.
61
Xu
,
A.
,
R. L.
Buchanan
, and
S.
Micallef
.
2016
.
Impact of mulches and growing season on indicator bacteria survival during lettuce cultivation
.
Int. J. Food Microbiol
.
224
:
28
39
.

Supplementary data