The 4-year study was performed in experimental fields at the ESAREC, located in a major tomato and vegetable producing area of Virginia. The popular commercial tomato cultivar ‘BHN 602’ (BHNSeed, Immokalee, FL, United States) was used for this study. Tomato seeds were sowed in potting mix (Premier Horticulture Inc., Quebec, QC, Canada) and grown in a greenhouse located on-site (ESAREC). The temperature in the greenhouse ranged from 23 to 33°C, with an average temperature of 28°C during transplant production. Approximately 7-week-old tomato seedlings were transplanted to experimental fields at the ESAREC. Tomato plants were cared for according to Virginia Cooperative Extension recommendations (i.e., pest management and fertilization, etc.) (Wilson et al., 2012).

Four field experiments were performed during the 2012 to 2015 growing seasons (Table 1 and Supplementary Table S1) to investigate the concentrations of recoverable Salmonella in tomato fields cultivated under different agricultural practices, which included use of two irrigation water sources (pond and well water), three fertilizers/soil amendments [PL, PL Ash, and a conventional fertilizer (Triple superphosphate, TSP)], and three production practices (SP, SW, and NW). Plastic mulch used in this study was a 0.03-mm-thick virtually impermeable polyethylene film (Berry Plastics Corp., Evansville, IN, United States). Irrigation was conducted by pumping water from the pond or well, both located on-site at the ESAREC. Plants were irrigated daily (∼ 6 L water/sub-plot) with either pond- or well-water through Aqua-Traxx drip tapes (TORO, Riverside, CA, United States) throughout the growing season, except on days with sufficient precipitation. Soil amendments/fertilizers were applied 1 week before transplanting by incorporation into soil to achieve recommended nutrition levels (Wilson et al., 2012). Twenty-five kilograms of fresh PL, 4 kg PL ash, or 2 kg TSP were applied to each corresponding sub-plot to provide a standard rate of 100 kg /hectare available P₂O₅ according to Virginia Cooperative Extension fertility recommendations for tomato (Maguire and Heckendorn, 2015). The average soil nitrogen concentrations (NO3-N) in experimental fields were 1.12, 0.59, and 4.84 mg/kg at 0–25, 25–50, and 50–75 cm depth, respectively. After fertilization and growing tomatoes in the field, soil nitrate rates would only be expected to change to 4.62, 2.66, and 2.22 mg/kg at 0–25, 25–50, and 50–75 cm depth, respectively, (unpublished data, available upon request).

Experiment 1 was conducted annually during the growing seasons from 2012 to 2014 using a split-plot design with irrigation water source as the main plot factor (pond and well) and production practices as the sub-plot factor (SP, SW, and NW), with each combination replicated 4 times (total of 24 sub-plots, Table 1 and Supplementary Table S1). Experiment 2 was arranged in a split-plot design with fertilization source as the main plot treatment (PL, PL Ash, and TSP) and production practices as the sub-plot factor. In 2012 and 2013 for experiment 2, two production practices (SP and NW) were used as sub-plot factors, with each fertilization/production practice combination replicated 4 times (total of 24 sub-plots). In 2014 and 2015 for experiment 2, three production practices (SP, SW, and NW) were examined, with each fertilization/production practice combination replicated 4 times (total of 36 sub-plots). Experiment 3 was conducted in 2014 and 2015 and arranged in a split-plot design with irrigation water source (pond or well) as the main plots and fertilizer source (PL or TSP) as the sub-plots, with each combination replicated 4 times (total of 16 sub-plots). Tomato plantings in experiment 3 were staked and covered with plastic mulch (SP), which is the typical production practice for tomato production on the ESV. In experiment 4, an attenuated, kanamycin-resistant Salmonella Newport strain SN#17 was obtained from the Microbiology division of the Center for Food Safety and Nutrition (CFSAN), FDA (Allard et al., 2014). The attenuated strain was incorporated into the tomato plots through injection into the existing irrigation drip tapes using a Chemilizer fixed ratio injector (CH9000-210, Hydro Systems, Cincinnati, OH, United States) in the middle of August one time at the blossoming stage. Approximately 2.5 L (10⁴ CFU/ml) bacterial solution was inoculated to each sub-plot (7.5 m²) through drip tape placed on the soil surface of tomato beds with emitters spaced every 30 cm along the tape. The total population of attenuated strain for each sub-plot approximated 2.5 × 10⁷ CFU. This field experiment was performed in a separated experimental plot in 2014 by following the biosafety standard operating protocols approved by the Institutional Biosafety Committee at Virginia Tech (Permit No.: IBC # 17-051). It was arranged in a randomized complete block design with four replications per different production practice (SP, SW, and NW) for a total of 12 Salmonella-inoculated sub-plots. The same number of plants grown in plots using three different production practices and all drip irrigated by non-inoculated well water were used as controls.

Each sub-plot was 12.5 m in length and 0.6 m in width and contained approximately 30 tomato plants. A 3 m border row between all sub-plots was set up to reduce the cross contamination between treatments.

A HOBO micro station (Onset Computer Corporation, Bourne, MA, United States) was set up to record the average temperature and total rainfall over the course of the 4-year study (monthly detection in 2012–2013, and weekly in 2014–2015).

Pond and well water used for irrigation at the ESAREC were sampled at the beginning of each month to detect native Salmonella populations from August 2012 to December 2013. In 2014 and 2015, these water sources were sampled weekly. At each sampling time, 30 L (monthly detection) or 4 L (weekly detection) water samples were collected and stored on ice in the field prior to transport to the laboratory for analysis. Well water was pumped from ground, while pond water was collected from the surface area near the irrigation water intake.

Salmonella concentrations in water samples were determined using a previously described MPN method (Luo et al., 2014, 2016). In brief, three volumes of water samples (2000/500/100 mL for monthly samples, and 500/100/10 mL for weekly samples) were added to equal volumes of double strength lactose broth (LB, BD Biosciences, San Jose, CA, United States) with 12 and 4 replicates, respectively, and incubated at 37°C overnight. A 1 mL aliquot of each culture was subsequently transferred into 9 mL Tetrathionate Broth (TT Broth, Dot scientific inc., Burton, MI, United States) for selective enrichment at 37°C overnight. TT broth cultures were then streaked for isolation onto Salmonella selective XLT4 (BD Biosciences, San Jose, CA, United States) agar plates and incubated at 37°C overnight. Presumptive colonies (color black) were confirmed by the cross-streaking method using CHROMagar^TM Salmonella plates (DRG International Inc., Springfield, NJ, United States). Salmonella concentrations (MPN/liter) were determined by using MPN calculator build 23, created by Mike Curiale. The limits of quantitation were 0.03 to 25 MPN/liter for monthly sample analyses and 0.41 to 140 MPN/liter for weekly sample analyses, respectively, as determined by the range of the MPN analysis. Up to four confirmed colonies from each positive plate (as determined by the cross-streaking method) were stored in 20% glycerol at a −80°C freezer.

Fresh PL and PL ash samples were analyzed for native Salmonella population via MPN analysis prior to usage in field experiments 2 and 3. In 2012, 2500 g composite samples of fresh PL were collected from each of the 14 houses of three broiler farms on the ESV (farms A, J, and D) for Salmonella detection before the field trial. In 2013, moist fresh PL collected from spaces under drinking-water pipes for flocks, and dry fresh PL collected from spaces between the drinking-water pipes were sampled from one broiler chicken farm on the ESV and processed for detection of Salmonella spp. PL ash composites were collected from four different factories located in Virginia and Pennsylvania and tested for Salmonella before field application. During 2014 and 2015, 500 kg of fresh PL (naturally contaminated with Salmonella) was collected from one broiler farm on ESV and used as soil amendment for fertilization in the field trials (Gu et al., 2015).

Similar to the water analyses, three volumes of PL samples (50, 10, and 1 g), each collected in octuplicate, were assessed for Salmonella population density as described above. The lower and upper detection limits of the MPN assay for the PL and PL ash were 2.1 to 2,100 MPN/kg, respectively. Up to four confirmed colonies from each positive plate were stored in 20% glycerol at a −80°C freezer.

Plant rhizosphere samples were collected monthly during the field experiments from 2012 to 2015 and processed for detection of Salmonella spp. At each sampling point, 10-g of soil was collected from 30 different areas of the same sub-plot using pre-sterilized soil probes (Oakfield Apparatus, Inc., Fond du Lac, WI, United States) and combined to form three composite samples (100 g/sample). A total of 12 samples (3 composite samples/sub-plot × 4 replications) were taken at each sampling time from each treatment in each trial of all four field experiments. 75% ethanol was used to disinfect soil probes between sampling.

Three portions of plant rhizosphere (50, 10, and 1 g) were included in the MPN scheme to assess the Salmonella population. The same plating and cross-streaking methods as mentioned above were used for Salmonella isolation and confirmation. Experiment 4 was an exception, which substituted trypticase soy agar (TSA) plates containing 50 μg/ml kanamycin for isolation of the attenuated Salmonella strain SN#17. The lower and upper detection limit for Salmonella was 1.4 to 2,500 MPN/kg, respectively, for the rhizosphere samples. Up to four confirmed colonies from each positive plate were stored in 20% glycerol in a −80°C freezer.

Similar to rhizosphere sampling, tomato leaf samples were collected monthly during the field experiments from 2012 to 2015 for Salmonella detection. Composite sampling was used in each treatment per each trial. At each sampling time, 10 g of plant leaves were randomly collected with sterilized scissors from the 30 plants in each sub-plot to obtain three 100 g composite leaf samples. Leaves in each composite sample were chopped with sterilized scissors and stomached by Stomacher 400 Circulator (Seward Laboratory Systems Inc., Davie, FL, United States) in the sterile pre-labeled BA6141/CLR closure bags (Seward), and then transferred to equal volumes of double-strength LB. The same plating and cross-streaking methods as mentioned above were performed for Salmonella isolation and confirmation. Experiment 4 was the exception, where TSA plates containing 50 μg/ml kanamycin were used to isolate the attenuated Salmonella strain SN#17. Up to four confirmed colonies from each positive plate were stored in 20% glycerol in a −80°C freezer.

To mimic the production of commercial tomato growers on ESV, tomato fruits were harvested twice from each field trial. At each sampling time, 25 mature red fruits were randomly collected from each sub-plot. Mature red fruits were removed and placed into a sanitized pre-labeled harvesting bin. One disinfected bin was used per sub-plot. A total of 200 fruits (25 fruits/sub-plot × 4 replications × 2 sampling times) were sampled for each treatment in each field trial for Salmonella detection. All the bins with harvested fruit were stored in a walk-in cold storage room at 12–15°C overnight prior to microbial analysis. For Salmonella detection, harvested fruits were placed in sterile pre-labeled BA6141/CLR closure bags (Seward) and stomached by Stomacher 400 Circulator in the bags to release the liquid exudate. In the 2012 field trials of experiments 1 and 2, and experiment 4 in 2014, each harvested fruit was tested individually for Salmonella presence. After the 1st year of individual fruit sampling (2012), to save labor and testing cost, fruits were pooled (3–5 fruits pooled in one sampling bag) for the remaining field trials of experiments 1–3 (2013–2015). Ten milliliters of fruit puree per bag was then added to equal volumes of double-strength LB and incubated at 37°C overnight. The same plating and cross-streaking methods as mentioned above were performed for Salmonella isolation and confirmation. TSA plates with 50 μg/ml kanamycin were used to recover attenuated Salmonella strain SN#17 in experiment 4. Up to four confirmed colonies from each positive plate were stored in 20% glycerol at a −80°C freezer. In total, 4,800 tomato fruits were harvested from plots applied with naturally contaminated sources (pond water and PL, experiments 1–3), and 600 fruits from inoculated plots (experiment 4) for Salmonella detections (Table 1 and Supplementary Table S1).

Up to two stored Salmonella isolates from each positive plate were selected for serotyping analysis using the CDC standard protocol for the molecular determination of Salmonella serotype (Centers for Disease Control and Prevention [CDC], 2009). Briefly, DNA from a pure culture was isolated using Instagene (BioRad, Hercules, CA, United States). Multiplex PCR was set up using Qiagen HotStar Master Mix (Qiagen) and 1 μL of DNA, and thermocycled under the following conditions: 95°C, 15 min.; 30 cycles of 94°C for 30 s, 48°C for 90 s, 72°C for 90 s; then 72°C for 10 min. DNA from the PCR reactions were then hybridized to the beads (xMAP^® Salmonella serotyping assay, Luminex) containing specific O- and H-Ag probes before the addition of strepavidin-R-phycoerythrin (Invitrogen div. Life Technologies, Grand Island, NY, United States). After incubation the samples were read using the Bio-Plex instrument (BioRad, Hercules, CA, United States). Positives were determined based on the ratio of signal to noise using a negative control (no template DNA) as a baseline. Serotype was determined based on which antigens are positive for each sample. S. enterica serovar Typhimurium strain ATCC 14028 was used as positive control.

With MPN analysis, a value of zero MPN was assigned to any samples under the detection limits. Upper limit values were given to any samples over the detection limits. Due to the low population densities of Salmonella in filed samples, MPN data were convert to MPN/kg or MPN/L for following analyses. The Mann–Whitney U-test, as a non-parametric test, was used to compare Salmonella concentrations in all experiments, except for when Student’s t-test was used to compare Salmonella MPN values and the contamination ratio in rhizosphere samples (as normal distribution) between pond water irrigated plots in experiment 1 and PL amended plots in experiment 2. Correlation of Salmonella MPN values between pond water and plant rhizosphere samples in experiment 1 was analyzed by Pearson correlation. The average temperature of the sampling month/week as well as the total rainfall 1 month/week prior to the monthly/weekly sampling point were calculated and compared with Salmonella prevalence data in irrigation pond water samples by correlation analysis. The interaction and effects of irrigation water source, fertilizer type, and/or production practice on the presence of Salmonella in tomato fields were analyzed by general mixed (GLIMMIX) model.

Statistical analysis was performed using SAS (SAS release 9.3, SAS Institute Inc., Cary, NC, United States). Except when stated otherwise, P-values of <0.05 were considered statistically significant.

There were temporal (monthly or weekly) differences for Salmonella occurrence and concentration in irrigation pond water (Figure 1). The prevalence of Salmonella in pond water was 64.7% from August 2012 to December 2013 (11 of 17 months, Figure 1A), 11.8% in 2014 (6 of 51 weeks, Figure 1B), and 12.2% in 2015 (6 of 49 weeks, Figure 1C). The average Salmonella population density in positive pond water samples from August 2012 to December 2013 was 0.77 ± 0.31 MPN/L, and the average Salmonella concentrations in 2014 and 2015 were 4.06 ± 1.86 MPN/L and 2.30 ± 1.10 MPN/L, respectively.

Salmonella was isolated from the irrigation pond water during every growing season in the 4-year study. The highest observed Salmonella populations in pond water during irrigation occurred in September in 2012 (0.54 MPN/L, 95% confidence interval (CI): 0.29–0.99 MPN/L) and 2013 (1.3 MPN/L, 95% CI: 0.64–2.8 MPN/L) field trials, and in August 2014 (9.9 MPN/L, 95% CI: 3.4–29 MPN/L) and 2015 (3.4 MPN/L, 95% CI: 1.3–9.3 MPN/L) trials. In general, Newport was identified to be the most prevalent serovar of Salmonella isolated from pond water samples (Supplementary Figure S1D). However, the proportion of Salmonella Newport of identified isolates collected from pond water at the ESAREC decreased from 60 to 28% from 2013 to 2015 during the sampling periods examined in this study (Supplementary Figures S1A–C).

There were no significant correlations between weather parameters (temperature and rainfall) and Salmonella occurrence or MPN values in irrigation pond water during the study (P > 0.05).

No Salmonella was isolated from irrigation well water in this study.

Salmonella MPN levels in fresh PL samples collected from three broiler farms A, D, and J in 2012 varied in the 14 tested chicken houses, ranging from non-detectable (<2.1 MPN/kg) to maximum detection limit of 2100 MPN/kg (Figure 2A). Newport (n = 27 isolates), Saintpaul (n = 18), and Typhimurium (n = 8) were the top three Salmonella serovars isolated from fresh PL samples (n = 70) (Supplementary Figure S2A). Additional fresh litter from chicken houses A3, A4, and A7 was collected as fertilizer for soil amendment and applied to the field in 2012, prior to transplant.

In 2013, moist and dry fresh PL samples were collected from broiler chicken farm D, and PL ash from four different factories for Salmonella detection (Figure 2B). There were no significant differences in Salmonella MPN values between moist and dry PL samples (P > 0.05) gathered from the same farm, although it should be noted that the % moisture was not analyzed and all samples were reported on a wet-weight basis. The average population density of Salmonella in sampled fresh PL ranged from 16 to 540 MPN/kg (wet weight basis, Figure 2B). No Salmonella was isolated from PL ash samples (below detection limit of 2.1 MPN/kg). PL from chicken houses D3, D6, and D7 (circled in Figure 2B) were collected for use as the soil amendment in the field experiment in 2013. Typhimurium was the only Salmonella serovar identified from the fresh PL samples used for soil amendment in 2013 (n = 40).

Fresh PL was collected from broiler farm D on ESV in 2014 and 2015 as soil amendment for fertilization in the according field trials. Salmonella concentration in fresh PLs ranged from 34 MPN/kg (17–69 MPN/kg) in 2014 to 700 MPN/kg (270–1,800 MPN/kg) in 2015. Typhimurium (50%) and Kentucky (37%) were the two dominant Salmonella serovars isolated from PL in 2014 field trial (n = 39) (Supplementary Figure S2B), while serovars Typhimurium (45%), Newport (19%), Kentucky (16%), and Saintpaul (8%) were isolated from PL used for 2015 field trial (n = 49) (Supplementary Figure S2C).

No Salmonella was isolated from PL ash samples tested in this study.

In the 2012 field trial for plots irrigated with pond water, Salmonella population density and contamination ratio in the plant rhizosphere increased from August to September and then subsequently reduced for the duration of the trial (Figures 3A,B). The highest average Salmonella population densities in the rhizosphere were 24 (95% CI: 13–44), 13 (95% CI: 6.4–28), and 12 (95% CI: 5.4–25) MPN/kg under production practices of SP, NW, and SW, respectively. Salmonella population density and contamination proportion per month in rhizosphere were relatively higher in the plots covered with plastic mulch (SP) compared with the other two plant practices (NW and SW). Salmonella was isolated from the leaves of tomato plants in NW plots only in September with an 8.3% contamination ratio (1/12, Supplementary Table S2). Newport was the only Salmonella serovar identified from plant rhizosphere and leaf samples collected in the 2012 field trial.

In the 2013 field trial, Salmonella was only isolated from plant rhizosphere in the plots that were irrigated with pond water, and only in September, with average population densities of 6.4 (95% CI: 2.4–17 MPN/kg), 2.9 (95% CI: 0.74–12 MPN/kg), and 2.9 (95% CI: 0.74–12 MPN/kg) MPN/kg under production practices of SP, NW, and SW, respectively (Figures 3C,D). No Salmonella was isolated from the leaf samples in 2013. Salmonella serovars Newport, Typhimurium, and Kentucky were isolated from the rhizosphere samples in tomato plots irrigated by pond water in 2013.

In 2014, similar to the 2013 trial, Salmonella was isolated from plant rhizosphere samples collected in September from plots irrigated by pond water. The average population densities were 21 (95% CI: 11–39 MPN/kg), 9.8 (95% CI: 4.3–22 MPN/kg), and 9.8 (95% CI: 4.3–22 MPN/kg) MPN/kg under production practices of SP, NW, and SW, respectively (Figures 3E,F). Salmonella was also isolated from the leaves of the tomato plants in NW plots in September with a contamination ratio of 33% (4/12 leaves examined, Supplementary Table S2). Only Salmonella Newport was isolated from the samples collected from plots irrigated by pond water in 2014.

In all three field trials of experiment 1, Salmonella was not isolated from the well water or the plants or plant rhizosphere in well irrigated plots. There were, however, significant positive correlations between Salmonella population densities in pond water and plant rhizosphere samples from pond water irrigated plots (Table 2).

No Salmonella was isolated from the tomato fruits harvested in this experiment.

In the 2012 field trial, Salmonella population density and contamination ratio in plant rhizospheres of tomato plots, where PL was applied, decreased over time during the growth season from August to November. The initial average Salmonella population density in plant rhizospheres after soil amendment (with PL) was 240 MPN/kg (95% CI: 130–450 MPN/kg) under both production practices (Figures 4A,B). All the 12 plant rhizosphere samples tested in August and September in both SP and NW plots were Salmonella positive. Similar to experiment 1, Salmonella was isolated from the leaves of tomato plants in NW plots (leaves were in contact with soil) in September and October with 25 and 33% contaminated, respectively (Supplementary Table S3). Salmonella serovars Newport, Typhimurium, and Thompson were isolated from plant rhizosphere samples in the plots amended with PL, and Newport and Thompson were isolated from tomato leaf samples.

In the 2013 field trial, Salmonella was not isolated from the tomato fields (rhizosphere), leaf or fruit samples, even in the plots where PL was applied.

In the 2014 field trial, Salmonella was isolated from tomato plant rhizosphere that was treated with naturally contaminated PL. The initial average Salmonella population densities in plant rhizosphere samples were 30 (95% CI: 17–54 MPN/kg), 17 (95% CI: 8.8–34 MPN/kg), and 30 (95% CI: 17–54 MPN/kg) MPN/kg in July under production practices of SP, NW, and SW, respectively (Figures 4C,D). The initial contamination ratios were 75, 58.3, and 75% in SP, NW, and SW plots, respectively. Similar to the 2012 trial, Salmonella population density and contamination ratio in plant rhizosphere reduced during the growing season. Six of the 12 tomato leaf samples tested in NW plots were Salmonella positive in August (50%), and nine of the 12 NW leaf samples were positive in September (75%) (Supplementary Table S3). The top three Salmonella serovars isolated from plant rhizosphere samples in the plots amended with fresh PL were Kentucky (55%), Typhimurium (35%), and Newport (7%; n = 58). Only Kentucky and Typhimurium were isolated from tomato leaf samples in 2014 field trial.

In the 2015 field trial, Salmonella was only isolated from tomato plant rhizosphere samples in plots amended with fresh PL, with initial population density of 4.6–10 MPN/kg with 25–41.7% contaminated in July (Figures 4E,F). Similar to the observation in earlier trials, Salmonella population density and contamination ratio in plant rhizosphere decreased during the growing season. All tomato leaf samples tested in this trial were Salmonella negative. Typhimurium and Kentucky were the only two Salmonella serovars isolated from plant rhizosphere samples in the plots amended with PL in 2015 field trial.

In the four field trials of experiment 2, Salmonella was not isolated from plots fertilized with poultry liter ash or TSP. No contaminated fruits were detected in this experiment.

Based on the results derived from the six field trials of experiments 1 and 2, Salmonella population density and contamination ratio in plant rhizosphere samples were significantly higher in PL amended plots than that in pond water irrigated plots (p < 0.01).

In the 2014 field trial, Salmonella was isolated from nine of the 12 Pond+PL and Well+PL plant rhizosphere samples (75%) in July, with an average population density of 30 (95% CI: 17–54 MPN/kg) MPN/kg (Figure 5A). Salmonella population density and contamination ratio decreased under these two treatments during the growing season (Figures 5A,B). One of the 12 rhizosphere samples from Pond+TSP plots was Salmonella positive in August and September with a population density of 1.4 (95% CI: 0.2–10 MPN/kg) MPN/kg. Three of the 12 tomato leaf samples (25%) from Pond+PL plots were Salmonella positive in September (Supplementary Table S4). The major Salmonella serovars isolated from the plant rhizosphere samples were Kentucky (36%) and Saintpaul (26%) in Pond+PL plots (n = 58, Supplementary Figure S3A), and Kentucky (68%) and Typhimurium (26%) in Well+PL plots (n = 31, Supplementary Figure S3B). Newport and Saintpaul were the only two serovars isolated from plant rhizosphere samples in Pond+TSP plots, as well as contaminated leaf samples in Pond+PL plots.

In the 2015 field trial, Salmonella was only isolated from plant rhizosphere samples in Pond+PL, Well+PL and Pond+TSP plots. All the 12 Pond+PL samples (100%) and seven of the 12 Well+PL samples (58.3%) tested Salmonella positive in July, with population densities of 44 (95% CI: 25–77 MPN/kg) MPN/kg and 19 (95% CI: 10–37 MPN/kg) MPN/kg, respectively (Figure 5C). Similarly, Salmonella contamination ratio and population density decreased under these two treatments during the growth season (Figure 5D). One of the 12 rhizosphere samples from Pond+TSP plots was detected Salmonella positive from August to October with a population density of 1.4 (95% CI: 0.2–10 MPN/kg) MPN/kg. Different from the 2014 field trial, no tomato leaf samples were detected Salmonella positive in 2015. Salmonella serovars Kentucky and Typhimurium were isolated from plots fertilized with fresh PL (Pond+PL and Well+PL). Typhimurium and Newport were isolated from positive plant rhizosphere samples from Pond+TSP plots.

In the two field trials of experiment 3, no Salmonella was isolated from the plots irrigated by well water and applied with conventional fertilizer (Well+TSP). Salmonella was not isolated from the harvested tomato fruits in this experiment.

Average population density and contamination ratio of Salmonella in plant rhizosphere samples of the two trials were considerably higher in Pond+PL plots than that in Pond+TSP plots (P < 0.01), while the difference between Pond+PL and Well+PL samples was not significant (P = 0.34). Salmonella levels in Well+PL plots was relatively higher than that in Pond+TSP plots (P = 0.09). Salmonella population density and contamination ratio in plant rhizosphere samples of all the above three treatments (Pond+PL, Pond+TSP, and Well+PL) were significantly higher than that in Well+TSP plots, where no Salmonella was detected (P < 0.05).

After artificial inoculation in the 2014 field trail, plant rhizosphere soils in all inoculated plots were contaminated by the attenuated Salmonella Newport strain SN#17 with an initial average population density of 160 (95% CI: 87–310 MPN/kg) MPN/kg (0.16 MPN/g, 95% CI: 0.087–0.31 MPN/g) and showed a declining trend during the growing season (Figures 6A,B). Similar to the observation in experiments 1 and 2 (Figures 3, 4), Salmonella MPN values and contamination ratios in plant rhizosphere were relatively higher in the plots covered with plastic mulch (SP). For tomato leaf samples, Salmonella SN#17 was isolated from six of the 12 samples collected from NW plots (leaves may touch soil as not staked) in August, September, and October (50%). Four tomatoes harvested from two of the four NW sub-plots (tomatoes not staked and no plastic mulch was used, tomatoes were observed on field soil) tested Salmonella positive. However, Salmonella was not detected from the fruits harvested from the inoculated SP and SW plots. The average contamination ratio on/in fruits was less than 1% (4 of 600 tested tomatoes). No Salmonella was isolated from control plots irrigated by well water without inoculation. Serotyping results confirmed that all Salmonella isolates from this experiment belong to the inoculated strain, serovar Newport.