Survival of Twelve Pathogenic and Generic Escherichia coli Strains in Agricultural Soils as Influenced by Strain, Soil Type, Irrigation Regimen, and Soil Amendment

Abstract

Biological soil amendments of animal origin (BSAAO) play an important role in agriculture but can introduce pathogens into soils. Pathogen survival in soil is widely studied, but data are needed on the impacts of strain variability and field management practices. This study monitored the population of 12 Escherichia coli strains (generic, O157, and non-O157) in soils while evaluating the interactions of soil type, irrigation regimen, and soil amendment in three independent, greenhouse-based, randomized complete block design trials. Each E. coli strain (4–5 log₁₀ CFU/g) was homogenized in bovine manure amended or nonamended sandy-loam or clay-loam soil. E. coli was enumerated in 25 g samples on 0, 0.167 (4 h), 1, 2, 4, 7, 10, 14, 21, 28, 56, 84, 112, 168, 210, 252, and 336 days postinoculation (dpi). Regression analyses were developed to understand the impact of strain, soil type, irrigation regimen, and soil amendment on inactivation rates. E. coli survived for 112 to 336 dpi depending on the treatment combination. Pathogenic and generic E. coli survived 46 days [95% Confidence interval (CI) = 20.85, 64.72; p = 0.001] longer in soils irrigated weekly compared to daily and 146 days (CI = 114.50, 184.50; p < 0.001) longer in amended soils compared to unamended soils. Pathogenic E. coli strains were nondetectable 69 days (CI = 39.58, 98.66, p = 0.015) earlier than generic E. coli strains. E. coli inactivation rates demonstrated a tri-phasic pattern, with breakpoints at 26 dpi (CI = 22.3, 29.2) and 130 dpi (CI = 121.0, 138.1). The study findings demonstrate that using bovine manure as BSAAO in soil enhances E. coli survival, regardless of strain, and adequate food safety practices are needed to reduce the risk of crop contamination. The findings of this study contribute data on E. coli concentrations in amended soils to assist stakeholders and regulators in making risk-based decisions on time intervals between the application of BSAAO and the production and harvest of fruits and vegetables.

Biological soil amendments of animal origin (BSAAO), including raw manure and poultry litter, play an important role in managing soil quality and fertility and improving nutrient content in both conventional and organic agriculture (Mikha et al., 2014; Rees et al., 2014; Lwin et al., 2018; Bello et al., 2021). While BSAAOs provide benefits to crop production in fields, the use of untreated or improperly treated BSAAOs may introduce human health risks to production environments, as untreated BSAAOs can harbor enteric pathogens (e.g., pathogenic Escherichia coli, Salmonella enterica) (Hutchison et al., 2005; Franz et al., 2007; Baker et al., 2019; Chen et al., 2019; Gu et al., 2019; Dunn et al., 2022) that can transfer to produce surfaces (Islam et al., 2004; Mootian et al., 2009; Jacobsen & Bech, 2012). As many growers rely on the application of BSAAOs to maintain soil health, the termination of the use of BSAAOs is not practical or feasible. Since fresh produce is often consumed raw or with minimal processing, good agricultural practices (GAPs) and other practices are applied during the growing and harvesting of fresh fruits and vegetables to minimize the potential for microbial contamination.

Large multistate outbreaks of E. coli O157:H7 have been linked to the consumption of contaminated fresh produce, causing a substantial economic and human health burden (FDA, 2019a, 2019b; CDC, 2022a, 2022b). As a result, agricultural inputs to fields, including BSAAO application, have been scrutinized during produce production. Research has shown that E. coli O157:H7 can be present in and may survive for extended periods in soils amended with BSAAOs (Islam et al., 2004; Franz et al., 2005, 2008; Sharma et al., 2016, 2019; Jay-Russell et al., 2018). For example, E. coli O157:H7 persisted in manure-amended soils for up to 217 days (Islam et al., 2004). Additional studies have shown environmental conditions (e.g., temperature), soil characteristics (e.g., soil composition), and management practices (e.g., irrigation) influence the survival and persistence of E. coli O157:H7 in amended soils (Franz et al., 2005, 2008; Semenov et al., 2008; Ongeng et al., 2011; Biswas et al., 2016). These studies typically examine a single strain or cocktail of E. coli O157:H7 strains; however, the survival of non-O157 pathogenic strains (e.g., O121, O145) received less attention. Shiga-toxin-producing E. coli (STEC) and enterohemorrhagic E. coli (EHEC) strains have been previously detected in amended soils (Baker et al., 2019; Ramos et al., 2021), and non-O157 STEC strains have been previously linked to produce-borne outbreaks (CDC, 2010, 2014; Luna-Gierke et al., 2014). A field survey conducted across four states (California, Minnesota, Maine, and Maryland) found the prevalence of non-O157 STEC [7.7% (190/2461)] was greater than that of E. coli O157 [0.04% (1/2461)] in soil amended with untreated manure (Ramos et al., 2021). There is a need to understand how a variety of pathogenic and generic E. coli strains behave under different treatments in amended soils. Currently, risk management strategies used to minimize microbial contamination of crops from BSAAOs focus on time intervals between the application of the amendment and harvest of the crop (USDA, 2010). For example, the National Organic Program (NOP) standards recommend waiting 120 days between the incorporation of untreated animal manures into soil and the harvest of crops when the edible portion of the crop comes in contact with the soil (USDA, 2010). Until the US Food and Drug Administration's (FDA) research and risk assessment is completed (USDA, 2010; FDA, 2015, 2018), the FDA Food Safety Modernization Act’s (FSMA) Produce Safety Rule (PSR) currently recommends maximizing this time interval and does not take exception to following the NOP time intervals between application and harvest for the use of untreated BSAAO on farms. Research examining growers' perspectives on BSAAOs through workshops, focus groups, and surveys revealed growers unanimously agreed that there is a need for more science-based data on the survival of foodborne pathogens in amended soils to better determine appropriate time intervals between incorporation of untreated BSAAOs to fields and harvest (Ramos et al., 2019). Data collected on the survival of different pathogenic and nonpathogenic strains in amended soils will fill data gaps and assist risk assessment efforts, allow for science-based policy decisions, and potentially provide beneficial information to growers. This greenhouse study was performed to examine the survival and inactivation of twelve E. coli strains (representing nonpathogenic generic, O157:H7, and non-O157 pathogenic E. coli) in agricultural soils, and evaluate how soil type (sandy-loam, clay-loam), irrigation regimen (daily, weekly), and soil amendment (unamended, bovine manure- amended) affected pathogenic and generic E. coli concentrations over time.

Materials and methods

Experimental design

This study was conducted in a greenhouse to safely employ pathogens in a controlled environment, following the design and setup previously described in Bardsley et al. (2021). Briefly, this study consisted of three independent randomized complete block design trials. Twenty-four pots (twelve sandy-loam soil and twelve clay-loam soil) were amended with bovine manure, inoculated with one of twelve E. coli strains, and irrigated daily.

Additionally, to understand the impact of irrigation regimen on survival, three E. coli strains were selected and inoculated (one strain per pot) into six bovine manure-amended pots (three sandy-loam soil and three clay-loam soil) that were irrigated weekly. The three strains were chosen to include one each for STEC O157, STEC non-O157, and non-pathogenic generic E. coli and various origins (e.g., environmental, animal, food). Furthermore, the generic E.coli strain was inoculated into two nonamended pots (one sandy-loam soil and one clay-loam soil to evaluate soil amendment use). The present study included a total of 32 pots per trial [twenty-four amended pots irrigated daily (twelve per soil type), six amended pots irrigated weekly (three per soil type), and two nonamended pots irrigated daily (one per soil type), with 96 pots across the three trials. For each collection time point, soil samples were collected in triplicate. Throughout the study duration, the greenhouses were maintained to mimic air temperature (20–27 °C) and relative humidity (55–65%) conditions expected during the late spring growing season in Virginia, USA. Greenhouse conditions were monitored continuously, and air temperature and relative humidity were recorded during each sample collection timepoint (HOBO Onset Micro Station Data loggers; Bourne, MA, USA; Fig. S1).

Pathogenic and generic E. coli strain preparation

The inoculum was prepared using a modified version of the protocol described by Sharma et al. (2016). The E. coli strains used in this study were selected to include a diversity of strains, including O157, various non-O157 pathogenic, and nonpathogenic generic E. coli strains as well as isolation sources (animal-, clinical-, environmental-, and food-associated sources; Table 1). Eleven individual strains and one three-strain cocktail of generic E. coli TVS (TVS 353, 354, and 355) were employed in the present study. The E. coli TVS cocktail was selected for use due to its frequent application in previous field and greenhouse studies as a nonpathogenic surrogate cocktail (Gutiérrez-Rodríguez et al., 2012; Sharma et al., 2016; Neher et al., 2019; Sharma et al., 2019; Pang et al., 2020). In order to distinguish between background microbiota and the inoculum, all E. coli strains were adapted to grow in the presence of 80 μg/mL rifampicin (R), as previously described (Parnell et al., 2005). Each strain was streaked from frozen stock culture (−80 °C) onto tryptic soy agar that was supplemented with 80 μg/mL rifampicin (TSA-R; Difco, Sparks, MD, USA) and incubated for 24 h at 37 °C. Using a 10 μL loop, 3–5 colonies of each strain were each added to 200 mL of tryptic soy broth supplemented with 80 μg/mL rifampicin (TSB-R; Difco, Sparks, MD, USA) and incubated overnight (18 h) at 37 °C with agitation (140 rpm). A concentration of 6 log₁₀ CFU/mL for each strain was prepared by diluting a 10 mL aliquot of each overnight culture into 990 mL of 0.1% peptone water (Difco: Sparks, MD, USA). Each E. coli strain population (inoculum) was confirmed by serial dilutions, spread plating onto TSA-R and MacConkey agar supplemented with 80 μg/mL rifampicin (MAC-R; Difco, Sparks, MD, USA), and incubated at 37 °C. After 24 h, colonies were counted and log₁₀ CFU/mL were calculated.

Table 1.

Description of the E. coli strains including source, and time to last E. coli detection (survival) in amended clay-loam and sandy-loam soils by daily and weekly irrigation regimen

Strain Letter	Strain Serotype Grouping	Source	Time (d) to Last Detect (E. coli Survival) and Overall Reduction (log10 MPN/g)a
			Bovine manure amended				Nonamended
			Daily Irrigationb		Weekly Irrigationc		Daily Irrigation
			Clay-loam	Sandy-loam	Clay-loam	Sandy-loam	Clay-loam	Sandy-loam
A	O157:H7	Food (Lettuce; provided by Dr. Manan Sharma)	252 (4.8–4.8)d	210 (5.3–5.6)
B	Generic	Environmental (three strain cocktail; TVS 353, 354, 355; provided by Dr. Trevor Suslow)	252 (4.8–4.9)d	252 (5.5–5.8)	336 (4.5–4.6)	336 (5.4–5.8)	112 (4.2–4.6)	112 (5.2–5.8)
C	O157:H7	Clinical (Human; PTVS 154; provided by Dr. Trevor Suslow)	168 (4.7–4.8)	210 (5.3–5.5)
D	O157:H7	Animal (Avian – Oregon junco; provided by Dr. Michael Cooley)	210 (4.6–4.7)	210 (5.2–5.6)	252 (4.6–4.8)	336 (5.5–5.7)
E	O157:H7	Environmental (Water; provided by Dr. Michael Cooley)	252 (4.6–4.7)	252 (5.6)e
F	Non-O157 Pathogenic	Environmental (Water; provided by Dr. Michael Cooley)	168 (4.6–4.7)	168 (5.0–5.7)
G	Non-O157 Pathogenic	Animal (Cattle; provided by Dr. Michael Cooley)	210 (4.6–4.8)	210 (5.3–5.6)
H	Non-O157 Pathogenic	Food (Lettuce; 2010 bagged romaine lettuce outbreak)	168 (4.6–4.7)	168 (5.4–5.6)	210 (4.7–4.8)	252 (5.4–5.7)
I	Generic	Environmental (Manure; provided by Dr. Manan Sharma)	252 (4.6–4.7)d	252 (5.3–5.5)
J	Generic	Environmental (Manure; provided by Dr. Manan Sharma)	252 (4.5–4.9)	252 (5.0–5.6)
K	O157:H7	Environmental (Manure; provided by Dr. Manan Sharma)	168 (4.4–4.5)d	168 (5.6–5.9)
L	Non-O157 Pathogenic	Environmental (Manure; provided by Dr. Manan Sharma)	168 (4.5–4.7)	168 (5.5–5.6)

^aValues outside parenthesis represent the longest days among three replicates postinoculation that E. coli was below the MPN limit of detection (−0.6 log₁₀ MPN/g) but was detectable in a 25 g soil sample. Values inside the parenthesis represent the overall log₁₀ reduction determined from the time of inoculation for the individual replicate and the level at the time to last detection (assuming to be −0.6 log₁₀ MPN/g).

^bAdjusting for the water loss daily.

^cAdjusting for the water loss weekly.

^dLog reduction value(s) not shown for one of three replicates, where the sample interval immediately preceding the “negative in 25 g” sample interval was above the MPN limit of detection.

^eFor this strain-soil combination, the time-to-last detection was available for one of three replicates.

Soil and bovine manure collection

Fresh bovine manure was obtained from a local cattle herd in Virginia on the morning of each of the three trials. Two soil types were used in the study presented here: sandy-loam and clay-loam. Sandy- loam soils were collected from produce fields in Painter, VA, while clay-loam was obtained from produce fields in Petersburg, VA. Prior to each trial, the soil was sieved to remove debris and break up aggregates and air dried for 7 days. Soil and bovine manure physicochemical characteristics (e.g., moisture, pH, total nitrogen) were determined by Waypoint Analytical (Richmond, VA, USA) and can be seen in Table S5. To understand the level of background microflora, for each of the three trials, bovine manure, and soil (25 g) were tested for rifampicin-resistant background microflora according to the protocols described below.

Soil inoculation and irrigation regime

For each of the 32 treatments, 100 mL of one of the E. coli was mixed with 1000 g of soil and 5 g of bovine manure (rate of application: 8.9 Mg/ha or 4 tons per acre). In a sterile plastic bag, soil was thoroughly homogenized by using alternating 30 sec shaking and 30 sec of rubbing for 5 min. Inoculated soil (∼4–5 log₁₀ CFU/g) was added to sterile 1.89L plastic pots, and sterilized distilled water was added to achieve the target moisture levels of 12.5% in sandy-loam and 22% in clay-loam soils (approximately 25 and 120 mL for sandy-loam and clay-loam soils, respectively). Following homogenization and irrigation, each pot was transferred to the greenhouse. Six control pots were also run in parallel and included noninoculated, bovine-manure amended, daily and weekly irrigated sandy-loam and clay-loam soils as well as noninoculated, nonamended, daily irrigated sandy-loam and clay-loam soils.

During irrigation events, soil moisture was maintained at 12.5 ± 2% and 22 ± 2% in sandy-loam and clay-loam soils, respectively, using a soil moisture probe (TDR150; Spectrum: Aurora, IL, USA). Soil moisture levels and irrigation regimens were chosen to mimic practices within the Southeastern regions, specifically the Mid-Atlantic, based on the Southeastern USA Vegetable Crop Handbook (Reiter, 2020) and the Mid-Atlantic Commercial Vegetable Production Guide (Kuhar et al., 2020).

Soil sampling and enumeration

E. coli concentrations in 25 g of soil were enumerated at 0, 0.167 (4 h), 1, 2, 4, 7, 10, 14, 21, 28, 56, 84, 112, 168, 210, 252, and 336 days postinoculation (dpi), or until E. coli was no longer detectable in two consecutive sampling events. For enumeration, a composite soil sample consisting of five 5-g samples was collected from each pot and added to a sterile, filtered Whirl-pak bag (Nasco: Modesto, CA, USA) containing 225 mL of buffered peptone water (BPW; Difco: BD, USA). Each Whirl-pak was hand massaged for 60 sec to homogenize. The homogenate was serially diluted in 0.1% peptone water (1:10), and 0.1 mL of each dilution was spread-plated, in duplicate, on MAC-R. To increase the limit of detection (LOD) to 1.0 log₁₀ CFU/g, 0.25 mL of homogenate was plated onto four MAC-R agar plates (1 mL total) in duplicate. All MAC-R plates were incubated for 24 h at 37 °C.

When E. coli counts approached the direct plating LOD, enumeration was performed using a most probable number (MPN) method to increase the LOD to −0.6 log₁₀ MPN/g (Jay-Russell et al., 2018; Gartley et al., 2018; Murphy et al., 2023; FDA, 2023). For MPNs, 5 mL of homogenate was transferred into the first eight wells of a 48-well reservoir (8 × 6 wells) with the following five rows of eight wells filled with 4.5 mL of tryptic soy broth (TSB; Difco, Sparks, MD, USA). Samples were serially diluted (1:10) five times and incubated for 2 h at 25 °C followed by 22 h at 42 °C with agitation (50 rpm). After incubation, a 50 μL aliquot from each of the 48 wells was transferred into the corresponding wells of a 48-well reservoir containing 5 mL of Modified Enterohemorrhagic Escherichia coli Broth (mEHEC; Millipore Sigma: Burlington, MA, USA). After incubating for 24 h at 42 °C with agitation, 10 μL from each well was channel streaked onto Chromagar STEC (CHROMagar: Paris, France) and incubated at 37 °C for 24 h. The amount of E. coli-positive streaks (out of eight) was documented, and MPN values were calculated. Concurrent with MPN enumeration, the remainder of the sample was enriched following the FDA Bacteriological Analytical Manual (BAM) methods (FDA, 2023).

Statistical analysis

All analyses were performed in R version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria). To understand the impact of soil-type, strain, and management practices on days to last detection (i.e., the last time point when E. coli could be isolated from a 25 g sample), a linear model was developed for the log₁₀ CFU/g or log₁₀ MPN/g data. The linear model included trials (1, 2, and 3), soil type (sandy- loam and clay-loam), bovine-manure amended (yes or no), irrigation regimen (daily or weekly), strain source (animal, clinical, cocktail, environmental, or food), and pathogenicity [yes (O157 and pathogenic non-O157 E. coli strains) or no (nonpathogenic generic E. coli strains] as fixed effects. The effect estimates of the linear model can be interpreted as the difference in total days E. coli was detectable due to a change in reference level. Separately, a multivariate linear regression model was developed to understand the daily inactivation rate of E. coli and included trial, soil type, bovine-manure amended, irrigation regimen, strain source, and pathogenicity. The effect estimated by the multivariate linear regression model can be interpreted as the change in log₁₀ daily inactivation rate (log₁₀ CFU/g or log₁₀ MPN/g) E. coli due to a change in reference level. Moreover, a Davies test was used to determine whether the multivariate linear regression model employed included a nonconstant regression parameter (the breakpoint). Since the Davies test was significant for two breakpoints, a segmented linear regression model was built to understand the daily inactivation rate (log₁₀ CFU/g or log₁₀ MPN/g). Furthermore, to understand the inactivation pattern for each strain, soil type, amendment, and irrigation regime combination, a Davies test was also employed for each combination separately to determine if and how many breakpoints were significant. If no breakpoints were identified, a linear regression model was built to understand the daily inactivation rate (log₁₀ CFU/g or log₁₀ MPN/g). If significant breakpoints were found, a segmented linear regression model was built to understand the daily inactivation rate (log₁₀ CFU/g or log₁₀ MPN/g).

Conditional inference trees were used to identify and visualize specific combinations of the full dataset that were associated with decreased or increased E. coli survival time. Conditional inference trees were built using the mlr and partykit packages in R (Hothorn & Zeileis, 2014; Bischl et al., 2016). To prevent overfitting, the mincriterion (the value of the test statistic needed for a split) was set to 0.95 while the maxdepth (maximum depth of the tree) was set to 3.

Results

Pathogenic and generic E. coli survival (time to last detect) and inactivation

No rifampicin-resistant background microflora were detected in soil or bovine manure prior to use. Given an initial level of 4–5 log₁₀ CFU/g, pathogenic and generic E. coli survival (LOD: −0.6 log₁₀ MPN/g) ranged from 112 to 336 dpi depending on strain, irrigation regimen, and soil type (Table 1, Fig. S2). The generic E. coli TVS strain cocktail (B) in nonamended soils demonstrated the shortest time to be undetectable, surviving 112 dpi in both sandy-loam and clay-loam soils. Conversely, the generic E. coli TVS strain cocktail (B) in both amended sandy-loam and clay-loam soils and irrigated weekly survived 336 dpi. Strain D (O157:H7) in amended sandy-loam soils and irrigated weekly also survived 336 dpi. Overall, the six strain and soil type combinations in amended soils that underwent weekly irrigation had prolonged survival, compared to the same strains in amended soils that underwent daily irrigation (paired comparison; Table 1). The overall E. coli reduction at the time to last detect, for the amended soils irrigated daily, ranged from 4.4 to 4.9 log₁₀ for clay-loam soil and from 5.0 to 5.9 log₁₀ for sandy-loam soils (Table 1). The log reduction was calculated assuming the level was −0.6 log₁₀ at the time of to last detect. For strains A and K (O157:H7) and strains B and I (generic E. coli) in clay-loam soils, log reduction was not shown for one of three replicates, where the sample-interval immediately preceding the “negative in 25 g” sample interval was above the MPN limit of detection. For strain E (O157:H7) in sandy-loam soil with daily irrigation, the time to last detect was 168 dpi (5.6 log₁₀ reduction) for one of three replicates; for the other two replicates, the time to last detect was not available. Rather, for one replicate, the level was 0.8 log₁₀ at 168 dpi and negative in 25 g at 210 dpi; and for the third replicate, the level was 0.5 log₁₀ at 252 dpi and negative in 25 g at 336 dpi. We used 252 dpi as the time for last detect for comparison with other strain-soil combinations. Pathogenic and generic E. coli concentrations (log₁₀ CFU/g or log₁₀ MPN/g; or negative in 25 g) at each time point for each strain by soil type, amendment, and irrigation regimen can be found in Tables S1 and S2.

The inactivation for all E. coli strains combined demonstrated a tri-phasic pattern, with two breakpoints: 26 dpi [95% Confidence Interval (CI) = 22.3, 29.2] and 130 dpi (CI = 121.0, 138.1). From 0 dpi (inoculation) up until 26 dpi, the quickest daily inactivation of E. coli was 0.07 log₁₀ CFU/g/day (CI = 0.07, 0.08). Following the first breakpoint, the daily inactivation of E. coli strains in soils was reduced to 0.03 log₁₀ CFU/g/day (CI = 0.02, 0.03) from 26 to 130 dpi. The slowest daily inactivation of E. coli strains was observed following the second breakpoint (>130 dpi) at 0.003 log₁₀ CFU/g/day (CI = 0.003, 0.004).

While a tri-phasic inactivation was observed when all data were considered, the inactivation and the number of breakpoints varied when data were separated by E. coli strain, soil type, amendment, and irrigation regimen (Table 2 and Tables S3 and S4). In the amended sandy-loam soil irrigated daily, strains A, F, G, H, J, and L had a tri- phasic inactivation, while strains C, D, E, I, and K had a biphasic inactivation. Only the generic E. coli TVS strain cocktail (B) had a linear inactivation (Table 2). All the non-O157 pathogenic E. coli strains (F, G, H, J, and L) demonstrated tri-phasic inactivation, while all but one of the O157:H7 E. coli strains (A) followed a biphasic inactivation in sandy-loam soils (C, D, E, and K). However, this inactivation trend was not observed in the amended clay-loam soils irrigated daily (Table 2). In amended clay-loam soils, E. coli strains D and E followed a tri-phasic inactivation, while E. coli strains A, B, C, F, G, H, I, J, K, and L demonstrated a biphasic inactivation (Table 2). All E. coli strains, except C and I, followed different inactivation patterns between amended sandy-loam and clay-loam soils that were irrigated daily.

Table 2.

Results of a segmented linear regression model developed to best characterize the inactivation of E. coli by strain and soil type for amended soils irrigated daily

Strain	Sandy-loam					Clay-Loam
	Breakpoint (days)a		Effect Estimate (die off rate)b		p value	Breakpoint (days)		Effect Estimate (die off rate)		P value
A					<0.001					<0.001
	0		−0.187	−0.216, −0.157		0		−0.019	−0.021, −0.018
	16.1	12.9, 19.2	−0.031	−0.039, −0.023		237.4	204.3, 270.5	−2.024 * 10−10	−0.008, 0.008
	93.8	75.5, 112.2	−0.001	−0.004, 0.001		–	–	–	–
B					<0.001					<0.001
	–	–	−0.020	−0.021, −0.018		0		−0.031	−0.039, −0.023
	–	–	–	–		72.1	42.8, 101.5	−0.011	−0.013, −0.008
C					<0.001					<0.001
	0		−0.044	−0.049, −0.039		0		−0.057	−0.065, −0.050
	123.7	100.9, 146.6	−0.003	−0.010, 0.004		73.1	62.7, 83.5	−0.002	−0.005, 0.001
D					<0.001					<0.001
	0		−0.143	−0.160, −0.126		0		−0.112	−0.142, −0.082
	32.1	28.2, 36.1	−0.004	−0.007, −0.002		16.8	11.4, 22.2	−0.023	−0.028, −0.017
	–	–	–	–		142.8	115.0, 170.6	−2.99 * 10−11	−0.004, 0.004
E					0.016					<0.001
	0		−0.106	−0.150, −0.061		0		−0.118	−0.146, −0.090
	17.8	10.1, 25.4	−0.014	−0.017, −0.012		16.4	11.8, 21.1	−0.022	−0.027, −0.017
	–	–	–	–		142.1	115.1, 169.1	−6.770 * 10−10	−0.004, 0.004
F					<0.001					<0.001
	0		−0.062	−0.122, −0.002		0		−0.062	−0.071, −0.053
	14.0	−12.5, 40.6	−0.034	−0.040, −0.027		71.6	60.6, 82.7	−0.003	−0.007, 0.001
	150.0	121.6, 178.4	0.001	−0.012, 0.013		–	–	–	–
G					<0.001					<0.001
	0		−0.061	−0.075, −0.047		0		−0.033	−0.037, −0.029
	28.0	13.4, 42.6	−0.025	−0.030, −0.020		129.147	107.2, 151.1	−0.002	−0.007, 0.003
	175.4	149.1, 201.7	−7.759 * 10−5	−0.006, 0.006		–	–	–	–
H					<0.001					<0.001
	0		−0.138	−0.170, −0.106		0		−0.045	−0.049, −0.040
	16.4	11.7, 21.2	−0.030	−0.036, −0.025		101.0	89.1, 113.0	−0.002	−0.006, 0.002
	125.9	100.1, 151.7	−3.945 * 10−10	−0.007, 0.007		–	–	–	–
I					<0.001					<0.001
	0		−0.025	−0.027, −0.023		0		−0.034	−0.038, −0.031
	214.5	178.3, 250.7	0.111	−0.009, 0.009		116.3	98.0, 134.7	−0.004	−0.007, −0.001
J					<0.001					<0.001
	0		−0.135	−0.160, −0.109		0		−0.03	−0.041, −0.031
	21.6	15.9, 27.4	−0.021	−0.030, −0.013		114.4	87.78, 141.1	−0.003	−0.009, 0.003
	155.4	113.4, 197.4	−0.001	−0.005, 0.004		–	–	–	–
K					<0.001					<0.001
	0		−0.125	−0.139, −0.112		0		−0.064	−0.075, −0.054
	45.6	40.9, 50.3	−0.001	−0.004, 0.002		61.87	49.7, 74.1	−0.004	−0.009, 0.001
L					<0.001					<0.001
	0		−0.087	−0.115, −0.058		0		−0.048	−0.052, −0.044
	17.7	8.3, 27.2	−0.036	−0.041, −0.031		96.7	85.8, 107.6	−4.3654 * 10−11	−0.004, 0.004
	130.1	110.4, 149.7	−1.436 * 10−10	−0.007, 0.007		–	–	–	–

^aThe breakpoint, in days for when the inactivation rates changes, and 95% confidence interval.

^bThe inactivation rate, and 95% confidence interval, represent the daily decrease in log₁₀ based on the breakpoints.

In amended sandy-loam soils that were irrigated weekly, all three strains (B, D, and H) demonstrated different inactivation patterns with the generic E. coli TVS strain cocktail (B) following a linear inactivation with no breakpoints, strain D (O157:H7) a biphasic inactivation, and strain H (non-O157 pathogenic) a tri-phasic inactivation (Table S3). In amended clay-loam soils that were irrigated weekly, the generic E. coli TVS strain cocktail (B) demonstrated a tri-phasic inactivation, while strains D (O157:H7) and H (non-O157 pathogenic) followed a biphasic inactivation (Table S3). As for the nonamended soils irrigated daily, the inactivation was only determined for the generic E. coli TVS strain cocktail (B), which had three breakpoints in sandy-loam soils (tri-phasic) and two breakpoints in clay-loam soils (biphasic; Table S4).

Impact of soil amendment, soil type, and irrigation regimen on pathogenic and generic E. coli concentrations

Regardless of strain, soil type, and irrigation regimen, pathogenic and generic E. coli in amended soils with bovine manure survived an average of 145.8 days longer (CI = 114.5, 184.5, p < 0.001), compared to generic E. coli TVS strain cocktail (B) in nonamended soils (Table 3). According to the multivariate linear regression model (inoculation to 226 dpi, without considering breakpoints), the daily inactivation rate of the generic E. coli TVS cocktail was 0.86 log₁₀ CFU/g/day (CI = 0.70, 1.02, p < 0.001) faster in nonamended soils, compared to amended soils (Table 4). Soil type was significant in both the linear model and the multivariate linear regression model, with amended clay-loam soils having a shorter survival or time to last detect (19.0 days shorter, CI = −32.7, −5.2, p = 0.009) and faster inactivation rate (0.64 log₁₀ CFU/g/day, CI = 0.57, 0.70, p < 0.001), compared to sandy-loam soils. Also, regardless of all other factors, E. coli survival (time to last detect) in amended soils that were irrigated weekly was 45.9 days (CI = 20.9, 64.7, p = 0.001) longer, compared to amended soils that were irrigated daily. Considering all strains collectively, no significant difference in inactivation rate (p = 0.323) was observed between amended soils irrigated on a daily versus weekly regimen; however, for certain combinations, the inactivation rate differed significantly between daily versus weekly regimen (e.g., strain D and strain H in amended sandy-loam soil).

Table 3.

Results of a linear model that were developed to investigate the impact of strain, soil type, irrigation regimen, and soil amendment on E. coli survival (time to last detect)

	Factor		Effect Estimatea	95% Confidence Interval	P value
Management Practices
	Irrigation Regime (Ref = Daily)		45.87	20.85, 64.72	0.001
	Soil Amendment (Ref = Amended)		−145.80	−184.50, −114.50	<0.001
Trial (Ref = First)
	Two		2.63	−14.20, 19.45	0.764
	Three		7.00	−9.82, 23.82	0.424
Soil Type (Ref = Clay-Loam)			19.00	5.22, 32.69	0.009
Strain Characteristics
	Pathogenic (Ref = No)b		−69.40	−98.66, −39.58	0.015
	Strain Source (Ref = Animal)
		Cocktail	40.41	10.74, 70.74	0.094
		Clinical	25.43	−21.74, 72.35	0.453
		Food	38.09	0.36, 72.54	0.182
		Environmental	−18.91	−66.07 28.02	0.572

^aSince all factors are categorical, the effect estimate can be interpreted as the difference in total days E. coli was detectable due to a change from the reference level to a different level.

^bIf serotype grouping (nonpathogenic generic strains, O157, and non-O157 pathogenic strains) was included instead of pathogenicity (yes vs no), there was a significant effect of serotype grouping. Specifically, E. coli O157:H7 and non-O157 pathogenic strains survived for 58.74 (95% CI = −86.47, −30.05; P = 0.048) and 84.36 (95% CI = −115.66, −53.20; P = 0.018), respectively, fewer days than nonpathogenic generic.

Table 4.

Results of a multivariate linear regression model that was developed to investigate the impact of strain, soil type, irrigation regimen, and soil amendment on daily inactivation rates of E. coli

	Factor		Effect Estimatea	95% Confidence Interval	P value
Management Practices
	Irrigation Regime (Ref = Daily)		0.05	−0.05, 0.14	0.323
	Soil Amendment (Ref = Amended)		−0.86	−1.02, −0.70	<0.001
Trial (Ref = First)c
	Two		0.02	−0.05, 0.10	0.545
	Three		0.15	−0.07, 0.23	<0.001
Soil Type (Ref = Clay-Loam)			0.64	0.57, 0.70	<0.001
Strain Characteristicsc
	Pathogenic (Ref = No)		−0.48	−0.56, −0.40	<0.001
	Strain Source (Ref = Animal)
		Cocktail	0.10	0.01, 0.19	0.030
		Clinical	0.25	0.10, 0.40	0.001
		Food	0.12	0.01, 0.22	0.030
		Environmental	0.17	0.02, 0.32	0.025

^aThe estimated can be interpreted as the difference in log₁₀ E. coli daily inactivation rate due to a change from the reference level to a different level.

Impact of E. coli strain on survival and inactivation

There was a significant effect of E. coli serotype grouping on survival (time to last detect), with nonpathogenic generic E. coli strains (B, I, and J) detectable for approximately 69 days (CI = −98.66, −39.58, p = 0.015) longer than O157:H7 and non-O157 strains (Table 3). When E. coli serotype grouping (nonpathogenic generic, O157, non-O157 pathogenic) was included in the linear model, instead of pathogenicity (yes or no), E. coli O157:H7 and non-O157 pathogenic strains survived for approximately 59 (CI = −86.47, −30.05; p = 0.048) and 84 (CI = −115.66, −53.20; p = 0.018) fewer days than generic E. coli, respectively (Table 3). Additionally, pathogenic E. coli strains died off at a significantly faster rate than generic E. coli (0.48 log₁₀ CFU/g/day, CI = 0.40, 0.58, p < 0.001). The length of E. coli survival (time to last detect) in amended soils was not significantly different among strain sources (Table 3), but daily inactivation rate of E. coli in soil did significantly differ based on strain source (Table 4).

Interaction between E. coli strain, soil type, and irrigation regimen

Findings from the conditional inference tree (Fig. 1) support the results of the linear model and the multivariate linear regression model. The algorithm selected irrigation regimen as the root node, confirming the substantial effect of irrigation regimen (daily versus weekly) on E. coli concentrations in amended soils. Weekly irrigation of amended soils resulted in the longest time that E. coli remained detectable, with no additional leaf nodes (i.e., no interaction with other variables). For daily irrigation of soils, the interaction between soil amendment (unamended, amended), type (generic, O157:H7, non-O157 pathogenic), and strain (12 E. coli strains) significantly influenced survival (time to last detect). The shortest survival (time to last detect) of E. coli in soil irrigated daily was in soils with no amendment (p = 0.002). Survival of E. coli in amended soils irrigated daily was split by generic or pathogenic E. coli (p < 0.001), and pathogenic E. coli further split by strain (p < 0.001) (Fig. 1). Interestingly, pathogenic E. coli strains did not split by O antigen or source.

Figure 1.

Conditional inference tree built showing potential interactions in the full dataset between irrigation regimen, soil amendment, and strain, and their impact on E. coli survival [time (days) to last detect].

Discussion

This study sought to understand and compare the survival and inactivation of pathogenic and generic E. coli in BSAAO-amended soils. Data generated here can help ongoing risk assessment efforts as well as develop potential risk mitigation strategies. While the fate of enteric bacteria in amended soils has been the subject of several previous research studies, this work aimed to investigate the variability in the survival of 12 E. coli strains and generate substantial data for modeling the range of inactivation patterns of pathogenic E. coli in risk assessment. This study generated inactivation patterns for each individual strain, where data were collected at sufficient intervals in the first two weeks and survival was monitored until the level was below the MPN limit of detection (−0.6 log10 MPN/g) and was negative in 25 g at two consecutive time intervals, providing complete inactivation for use in modeling. Additionally, this study was undertaken to examine the effect of soil type, irrigation regimen, and soil amendment on the different E. coli strains over time in a greenhouse setting. It should be noted that greenhouse conditions cannot truly mimic the diversity of field conditions observed in the US, but this limitation was balanced in this study by the benefit of using pathogenic strains versus surrogates or attenuated strains. Results from this study demonstrated generic E. coli strains, specifically the generic E. coli TVS cocktail (TVS 353, 354, and 355) survived substantially longer than the pathogenic E. coli strains and could be a useful surrogate for field studies. This aligns with other studies that have shown the TVS generic E. coli cocktail strains were more robust than other tested pathogenic E. coli strains (Busta et al., 2003; Eblen et al., 2005; Tomás-Callejas et al., 2011; Yin et al., 2020; Murphy et al., 2023).

E. coli survival in amended soils was dependent on strain and isolation source

Generic E. coli survived the longest and had the slowest inactivation rate in amended soils. Previous meta-regression analyses demonstrated similar results, finding that the average daily rate of decline for nonpathogenic E. coli (0.052 log₁₀ CFU/day) was significantly lower than that of pathogenic E. coli (0.071 log₁₀ CFU/day; p < 0.0001) in agricultural soils (Franz et al., 2014). Furthermore, previous work comparing nonpathogenic generic E. coli and attenuated E. coli O157:H7 survival in manure-amended soils showed that generic E. coli strains survived a total of 21 days longer, while the attenuated E. coli O157:H7 strain survived 7 days (Gutiérrez-Rodríguez et al., 2012). Since generic E. coli survived substantially longer than pathogenic E. coli in the present study (59 to 85 days for E. coli O157:H7 and non-O157 pathogenic strains, respectively) as well as in previous research, generic E. coli may overestimate pathogenic E. coli behavior in amended soils and lead to unnecessary or unintended food safety decisions. Results of the linear model showed E. coli O157:H7 strains survived longer on average than non-O157 pathogenic strains; however, this result was not significant based on 95% CIs. Previous studies support the impact of the soil environment on E. coli O157:H7 and non-O157 pathogenic survival (Ibekwe et al., 2014; Ma et al., 2014; Naganandhini et al., 2015). For example, Ma et al. (2014) noted that the time-to-last detection for three O157:H7 and three non-O157 strains in three agricultural soils from three major fresh produce growing areas (i.e., Yuma, AZ; Imperial Valley, CA; Salinas, CA) varied by strain and soil type with generally longer survival of non-O157 strains (16.5 ± 1.70 to 98.2 ± 1.06 days) compared to O157:H7 strains (19.2 ± 4.17 to 49.3 ± 3.75 days).

While evidence suggests considerable phenotypic diversity among foodborne pathogen strains from various sources (e.g., clinical, environment, food), studies often do not investigate the survival of organisms based on the source of isolation (Gutiérrez-Rodríguez et al., 2012; Franz et al., 2014; Naganandhini et al., 2015; Ma et al., 2014). The present study observed a significant difference in the survival capability of animal-, environmental-, food-, and human-associated E. coli isolates in manure-amended soil. Some laboratory-based studies also showed substantial variability in the survival of human- or animal- associated E. coli O157:H7 strains in cattle manure-amended soil (Franz et al., 2008; Sharma et al., 2019). Results demonstrated E. coli strains sourced from animals had the quickest inactivation rate in amended soils when irrigated daily, followed by food-, environmental-, and clinical- (human) associated E. coli strains. A previous study examining the survival of 18 E. coli O157:H7 strains in sandy soils amended with bovine manure found the average survival duration was significantly longer for human isolates (179 days) compared to animal isolates (98 days, p = 0.025) (Franz et al., 2011). The extended survival of clinical strains is not unexpected, as research has shown that foodborne pathogen strains from clinical-associated sources can survive and adapt to stressors better than strains isolated from environmental or food sources (Dykes & Moorhead, 2000; Vialette et al., 2003; Lianou & Koutsoumanis, 2013; Castro et al., 2019). These results highlight the importance of pathogenicity, serotype grouping, and isolation source in research studies aimed at evaluating the behavior of pathogenic bacteria for the development of risk assessments or risk management. However, the current study investigated and compared the survival of strains inoculated individually into amended soil, so future work may explore interactions of different E. coli strains when inoculated as a cocktail.

Management practices including irrigation regimen and soil amendment impacted pathogenic and generic E. coli survival in soil

Based on the conditional inference tree, irrigation regimen, and soil amendment were associated with E. coli survival (time to last detection) in soils. Also, irrigation regimen and soil amendment significantly influenced E. coli survival in the linear and multivariate linear regression model. E. coli survived significantly longer in amended soils with untreated bovine manure, as compared to nonamended soils (no incorporation of bovine manure to soil). This finding is consistent with prior research demonstrating that E. coli survives for extended periods of time in both bovine manure and amended soils with bovine manure (Jiang et al., 2002; Franz et al., 2005, 2008, 2011; Semenov et al., 2008; Sharma et al., 2016, 2019; Siriphap et al., 2017; Pang et al., 2020). This study, combined with the prior studies, supports the hypothesis that application of untreated BSAAO (e.g. manure) can facilitate E. coli survival. Enhanced survival of E. coli in bovine manure-amended soils may be attributed to high availability of nutrients (e.g., carbon, nitrogen, phosphorous) in bovine manure amendments (Mikha et al., 2014; Rees et al., 2014; Lwin et al., 2018; Sharma & Reynnells, 2018; Bello et al., 2021). The National Organic Program (NOP) standards recognize the added risk associated with untreated BSAAO use and recommend a 90- or 120-day waiting period between application to fields and harvest of crops (USDA, 2010); however, in this study, only the nonamended soils were undetectable by 120 days. All pathogenic E. coli strains tested in this study were detectable in both soil types amended with bovine manure from 168 to 336 days. While this may provide evidence that the NOP standards may not be sufficient to eliminate E. coli contamination risks under certain conditions for certain strains, it is difficult to ascertain the risk of foodborne illness without considering other factors like the initial level of contamination, environmental conditions in fields (e.g., UV, temperature extremes), horticultural/management practices (e.g., plastic mulch cover, staking of crops), competitive microflora, transfer coefficients to crops, among other factors. These factors can be considered in risk assessments when setting appropriate waiting periods between the application of untreated BSAAOs to fields and the harvest of crops. Additionally, since the present study utilized untreated bovine manure, future research incorporating treated manure, which does not require a waiting period under the NOP, would be valuable. This approach could offer further insights into the impact of manure treatment on microorganism dynamics.

In the study presented here, weekly irrigation of amended soils was associated with increased E. coli survival by 46 days compared to daily irrigation of soils. Previous research has shown soil moisture to enhance E. coli survival (Chandler & Craven, 1980; Cools et al., 2001; Underthun et al., 2018; Wang et al., 2018; Sharma et al., 2019). For example, when Sharma et al. (2019) analyzed data from twelve individual field trials to determine the survival duration of inoculated E. coli in manure-amended soils, soil moisture content on day-zero (day of manure application) accounted for the greatest percentage of variability attributed to any single factor in survival for generic E. coli (33%) and attenuated E. coli O157:H7 (29%). Additionally, previous work determined weather events influenced soil moisture, such as rainfall or flooding, and have found a significant impact on the survival of E. coli in agricultural soils (Busta et al., 2003; Underthun et al., 2018; Sharma et al., 2019; Litt et al., 2021).

Pathogenic and generic E. coli inactivation in soil is nonlinear, demonstrating a segmented inactivation with a faster initial inactivation

Overall, E. coli inactivation was tri-phasic with a faster initial decline followed by two additional periods of tailing. Of note, previous field studies (Moyne et al., 2011; Bezanson et al., 2012) show a 1–2 log reduction of E. coli O157:H7 population several hours after inoculation on lettuce plants under certain conditions. This study was designed to evaluate whether a rapid initial decline would occur in BSAAO- amended soils; however, such a decrease was not found in the population of the O157 and non-O157 pathogenic strains in amended clay- loam and sandy-loam soil 4 h after inoculation.

Of the 32 individual inactivation rates, only two [generic E. coli TVS strain cocktail (B) in amended sandy-loam soils irrigated (i) daily and (ii) weekly] did not follow a biphasic or tri-phasic inactivation. In all segmented linear regression models, the highest inactivation rate always occurred within the first breakpoint and was followed by slower inactivation rate(s) at each subsequent breakpoint. Several prior studies support these findings, which demonstrate E. coli inactivation in agricultural soils does not follow a linear inactivation (Franz et al., 2005; Semenov et al., 2007; Ongeng et al., 2011; Yao et al., 2015; Litt et al., 2021; Bardsley et al., 2022). For example, in a 2021 study by Litt et al. (2021), a nonlinear sigmoidal regression model accurately characterized E. coli survival in soil over 120 days; however, the number of sigmoids varied between the year and type of soil amendment, suggesting that temporal and agricultural practices influenced survival models. Likewise, in the study reported here, soil type contributed to survival models for the various E. coli strains and rarely followed a similar inactivation. Our findings, along with the prior studies, suggest a linear regression model may not accurately capture E. coli inactivation in amended soils under various environmental conditions. Instead, E. coli inactivation often exhibited curvilinear patterns. The utility of nonlinear regression models may be useful to explore as part of establishing E. coli inactivation rates in amended soils.

The use of untreated biological soil amendments of animal origin, like bovine manure, in agricultural soils can be both beneficial to crop production soils (e.g., soil fertility, soil health) and a food safety concern (potential introduction of pathogenic microorganisms). The current study observed variable survival and inactivation of E. coli in amended soils by strain, soil type, and irrigation regimen highlighting the challenges of managing BSAAO application across farms (each having unique characteristics). Given the complexities in managing E. coli in agricultural settings, future research utilizing probabilistic modeling, which can involve the integration of parameters in the form of distributions and account for uncertainty and variability in data and parameters, should be employed for an enhanced understanding of risk. Additionally, this study demonstrated a single linear inactivation rate is likely not adequate for characterizing pathogenic and generic E coli reductions in amended soils. However, a full factorial design was not used in this study, and the unbalanced evaluations potentially overlooked interactions that may have significantly impacted the survival of E. coli, which is a limitation of the present study. Nonetheless, the data presented here may be used in risk assessments and to aid growers, regulators, and other stakeholders in making decisions on time intervals between BSAAO application and harvest. Data on strain isolation sources (e.g., environmental, clinical, food) and model selection may also help develop future studies aimed at understanding pathogen survival and inactivation in produce preharvest environments, such as amended soils.

Data availability statement

Raw data are available upon request.