Loss of Virulence Genes in Escherichia coli Populations during Manure Storage on a Commercial Swine Farm

Abstract

Confined livestock production farms typically store their wastes prior to land application. Here, we employed three complementary approaches to evaluate changes in the population structure and stability of virulence genes in Escherichia coli during manure storage on a commercial farm that housed healthy swine. Isolates were genotyped by repetitive extragenic palindromic PCR using the BOXA1R primer and evaluated for the presence of selected virulence genes by PCR. Isolates obtained from the manure holding tank (n = 392) carried estB, fedA, stx_2e, astA, paa, aida-I, and sepA at lower frequencies than isolates obtained from fresh feces (n = 412). Fresh fecal material from the barn was added into diffusion chambers and immersed in the manure holding tank for 7 weeks. The fecal E. coli population was initially dominated by a single genotype, all isolates of which carried fedA and aida-I. After 7 weeks, a genotype that did not carry any virulence genes dominated the surviving population. In a second experiment, 48 fecal isolates of E. coli that varied in their genotypes and virulence gene complement were incubated in diffusion chambers in the manure holding tank for 3 weeks. Over 95% of the inoculum population carried at least one virulence gene, whereas after 3 weeks 90% of the recovered isolates carried no virulence genes. Taken together, these results indicate that during commercial manure storage, there was a significant reduction in the carriage of these virulence genes by E. coli. We propose that loss of virulence genes from enteric pathogens in the farm and in natural environments may, if generalized, contribute to the attenuation of a public health risk from contamination with agricultural wastes.

Escherichia coli is a commonly used indicator of fecal pollution of drinking and recreational water, and acceptable water microbiological quality standards are often determined on the basis of E. coli contamination (17, 20). E. coli is a ubiquitous commensal member of the gastrointestinal tract and is also an opportunistic pathogen for humans and animals (23). Numerous virulence factors have been identified in E. coli, and specific complements of virulence genes are associated with various pathotypes that cause diarrheal or extraintestinal diseases (23, 30). Virulence genes can also be carried by commensal E. coli (6, 15) and may have a role in promoting commensal attachment to intestinal epithelia (42). Many virulence genes have been found stably inserted within the chromosome in pathogenic E. coli clones (14, 26, 31). However, virulence genes can also be associated with elements that promote plasticity within the genome or exchange between bacteria, including plasmids, phages, integrons, and transposons (21, 28, 43, 50). The effect of virulence genes on fitness and survival in environmental matrices including manure, soil, and water is generally unknown. However, carriage of some virulence genes increases survival of E. coli in grazing protozoa and thus may enhance fitness in a secondary host widely dispersed in the environment (1, 49). Virulence factors that target attributes shared widely by eukaryotes may have a role in the defense against grazing protozoa and in pathogenicity and may thus promote environmental fitness (36).

Surface water and groundwater can potentially become contaminated with enteric bacteria in the effluent from land fertilized with manure or in leakage or spills from manure storage facilities (2, 46). Enteric pathogens of agricultural provenance generally must survive for extended periods of time in varied secondary habitats before they are ingested by humans and cause illness. For example, a typical human exposure pathway mediated through contaminated surface water or groundwater compromised by fecal material generated by a confined livestock operation would require the survival of pathogenic bacteria in manure storage for some period of time following application to the soil and in the water following contamination by runoff or leaching (45, 51). Exposure through ingestion of raw vegetables compromised by fecal material or contaminated irrigation water would, in addition to survival in these environmental matrices, also require survival on the surface of the crop in the field and through the retail chain (9). Virulence attenuation of some human-, animal-, and plant-pathogenic bacteria upon laboratory cultivation is a well-known phenomenon, and strains domesticated by in vitro propagation can differ substantially from their progenitors through mutation, deletion, plasmid curing, and genomic recombination (18, 22). But the genomic stability of pathogens in the natural environment and the potential impact of virulence attenuation on public health risks from agricultural effluents are generally unknown. Attenuation of virulence during the journey from the farm to the human consumer would reduce the risk from zoonotic pathogens (25). Furthermore, on-farm virulence attenuation would reduce the reservoir of bacteria able to propagate illness within a herd or between farms (10).

In many commercial production systems, animals and poultry are raised confined in barns, and their waste is collected and stored for several months before being used as fertilizer on agricultural land when the climate and crop conditions are suitable. In Canada, 85% of the swine are raised on farms where manure is stored as a slurry in static storage systems (3, 8). During storage the abundance of many pathogenic and nonpathogenic bacteria can decrease significantly, and the population composition can change profoundly (11, 29). Therefore, manure storage systems represent the first critical secondary habitat in which enteric bacteria must survive if they are to be released from these farming systems into the broader environment. In the work reported here, we determined if the frequency of carriage of selected genes conferring pathogenic potential in E. coli changed substantially during swine manure storage. We hypothesized that if a particular virulence gene enhanced fitness outside the host, it would be detected at a higher frequency in manure than in fresh fecal material; alternatively, if it carried a fitness cost, it would be detected at a lower frequency in manure. We obtained over a 6-month period from a single commercial swine farm approximately 2,500 E. coli isolates freshly shed by the animals and the same number of isolates from the farm's manure storage tank (16). By means of repetitive extragenic palindromic PCR (REP-PCR), we characterized the population structure, and by gene-specific PCR we determined the frequency and distribution of 12 virulence genes within the two populations. Six of the genes (fedA [48]), faeG [52], fanA [38], fasA [38], paa [32], and stx_2e [24]) are to date mostly associated with swine pathotypes, whereas six others (estA [30], estB [23], elt [30], aida-I [5], astA [30], and sepA [4]) are also associated with the virulence potential of bacteria in other host species including humans. Furthermore, using isolation chambers, we incubated fresh fecal material and defined mixtures of E. coli isolates within the manure holding tank to elucidate under more controlled conditions in situ changes in the population during storage.

MATERIALS AND METHODS

Characteristics of the swine farm.

The swine farm studied here has been described previously in Lu et al. (33). Briefly, it was a farrow-to-finish operation, with approximately 2,000 animals. Waste from the barn was stored as a slurry in an open concrete manure holding tank that had a capacity of about 800,000 liters. The tank was emptied in the spring and in the fall. The animals received a feed mix consisting of corn and soybean meal. During the course of the study described here, the animals received the following antibiotics. Nursery pigs received a growth promotion level of lincomycin-spectinomycin (Linco-spectrin), and finishing pigs received 40 g metric ton⁻¹ (40 ppm) of tylosin phosphate (Tylan) in their feed. Penicillin G was added to the water for 2 weeks after the animals were moved from the nursery to the finishing pens. Oxytetracycline was fed to the nursing sows (330 g metric ton⁻¹; 330 ppm) in January and April and to the dry sows (550 g metric ton⁻¹; 550 ppm) in April 2005. No therapeutic treatments were reported during or prior to the incubation experiments, and the herd was generally healthy throughout the study period.

Sampling of freshly shed feces and stored manure.

Freshly shed feces were collected from the floor of the swine barn, and stored manure was sampled from the manure holding tank monthly from March to August 2005. Procedures for sample collection, isolation, and confirmation of E. coli are described in Duriez and Topp (16). The distribution of selected virulence genes was determined as described below in a subset of the fecal (n = 2,193) and the stored manure (n = 2,475) collections described in detail in Duriez and Topp (16).

In situ incubations of E. coli in a commercial swine manure holding tank.

Diffusion chambers that provided contact of introduced materials with stored manure through 0.45-μm-pore-size polyvinylidene difluoride membranes (45-mm diameter; Millipore, Fisher Scientific, ON, Canada) were constructed. The chambers were designed to equilibrate introduced materials with the stored manure with respect to soluble organic and inorganic material, colicins, viruses, pH, and Eh, while preventing any exchange of bacteria. The chambers had an internal volume of 110 ml and were resistant to chemical disinfectants for surface sterilization. The membranes were protected from puncture during introduction and withdrawal from the stored manure by means of 10-cm-long flanges projecting from the chamber. Chambers were held at a depth of 0.5 m into the stored manure. Periodically, the chambers were removed from the manure holding tank to sample their contents. Sampling operations were undertaken on the farm. Chambers were thoroughly rinsed with water, surface sterilized by spraying with the disinfectant CiDecon (diluted 1:128; Decon Laboratories, VWR, Mississauga, ON, Canada), and after 5 min rinsed again with autoclaved water. The chambers were shaken to resuspend the slurry, and 1.5 ml of the contents was carefully removed by pipette and collected in 15-ml sterile tubes. Chambers were sampled, resealed, and reinserted into the manure holding tank within about 30 min.

Changes of an E. coli population during storage of fresh fecal material.

Two diffusion chambers were filled with slurry prepared with fresh fecal material collected from the barn in September 2006. At the time of sampling all pigs appeared healthy. Feces were collected from the floor of 12 pens of the swine barn. Approximately 5 g of each of the 12 fecal samples was thoroughly mixed, and the composite sample was resuspended in sodium metaphosphate buffer (2 g liter⁻¹) to give a 5% (wt/vol) suspension, approximating the same solid content as the stored manure. The chambers were immersed and incubated in the lagoon for 7 weeks, after which the manure holding tank was emptied.

For comparison with the in situ incubation of freshly shed slurry in the manure holding tank, parallel incubations were done in the laboratory as follows. Three autoclaved 1-liter Mason jars were filled with 800 ml of the same batch of fresh fecal slurry used in the in situ manure holding tank incubation, securely closed, and stored in the dark without agitation in the laboratory. Samples were taken periodically after the jars were shaken to homogenize the contents; the lids were removed and 1.5 ml of the contents was removed by pipette. The microcosms were sampled on the same days as the isolation chambers, and both experiments were stopped at the same time.

Changes of a reconstituted E. coli community in a manure storage holding tank.

Three diffusion chambers containing a mixture of characterized E. coli strains were incubated in the lagoon. In preliminary experiments, E. coli was found to be much less persistent in the manure supernatant or in autoclaved manure than in the stored manure slurry (data not shown). In order to create conditions that were axenic but that were as representative as possible of the stored manure, the E. coli isolates were incubated in chambers containing reconstituted sterile manure prepared by resuspending autoclaved manure solids that had been washed two times in the same volume of sterile sodium metaphosphate buffer. The autoclaved reconstituted manure was inoculated with a mixture of 48 different E. coli strains that had previously been isolated from fresh fecal material (in 2005) (16) and that captured a range of carriage of virulence genes, antibiotic resistance profiles, and REP-PCR-defined genotypes (Table 1). In order to minimize the nutritional shock that the cells experienced when introduced into the highly reduced manure, the isolates were grown anaerobically in Brewer's broth, prepared with tryptone (5 g·liter⁻¹; Difco, Fisher Scientific, Ottawa, ON, Canada), proteose peptone no. 3 (10 g·liter⁻¹; Difco), yeast extract (5 g·liter⁻¹; Difco), dextrose (10 g·liter⁻¹; Sigma-Aldrich Canada Ltd., Mississauga, ON, Canada), sodium chloride (5 g·liter⁻¹; Sigma-Aldrich Canada Ltd.), sodium thioglycollate (2 g·liter⁻¹; Sigma-Aldrich Canada Ltd.), and sodium formaldehyde sulfoxylate (1 g·liter⁻¹; Across Organics, Fisher Scientific), final pH 7.2. Inocula were grown statically for 24 h at 37°C in an anaerobic chamber (MGC; Fisher Scientific, Ottawa, ON, Canada) under an atmosphere of 20% CO₂ (AnaeroPack system; VWR, Mississauga, ON, Canada). The optical density at 600 nm of the cultures was measured, and enough cells were pipetted in the manure suspension to reach a cell density of 5 × 10⁵ bacteria ml⁻¹ for each of the 48 isolates (assuming that an optical density at 600 nm of 1 approximates 1 × 10⁹ bacteria ml⁻¹). The chambers were maintained in the manure tank until the population declined to approximately 1% of its starting value. Potential contamination of the chamber contents during the incubation was evaluated by plating slurry samples on m-Enterococcus agar (Difco, Fisher Scientific, Ottawa, ON, Canada), and the detection of enterococci was considered to be implicit evidence for contamination of the reconstituted autoclaved manure that otherwise should have contained only E. coli.

TABLE 1.

Virulence genes of strains used in the in situ incubation of a reconstituted E. coli community in a commercial manure holding tank

Genotype designationa	Isolate no.	Virulence genotypeb
282	1	fedA stx2e
	10	estB fedA aida-I astA paa
	11	estB fedA aida-I astA
	19	estB fedA aida-I
	22	estB aida-I astA
	23	estB fedA aida-I paa
	26	estB fedA aida-I stx2eastA paa
	27	estB fedA aida-I stx2eastA paa
	29	fedA astA
	31	estB fedA aida-I astA
	35	estB aida-I astA
	40	estB aida-I astA
	41	estB aida-I astA
	43	estB aida-I astA
	44	estB aida-I astA
	47	fedA paa
283	15	ND
	17	estB fedA aida-I astA paa
	25	estB fedA aida-I stx2eastA paa
	28	fedA stx2eastA paa
	30	estB fedA aida-I stx2eastA paa
	32	estB fedA aida-I stx2eastA paa
	33	estB fedA aida-I stx2eastA paa
	37	ND
	38	estB fedA aida-I stx2eastA paa
	39	estB fedA aida-I stx2eastA paa
	45	estB fedA aida-I stx2eastA paa
	46	fedA stx2eastA paa
	48	fedA stx2eastA paa
90	4	estB fedA aida-I astA
	5	estB fedA aida-I
	7	estB fedA aida-I
	13	estB fedA aida-I astA
	18	estB paa sepA
41	2	fedA aida-I
	34	estB fedA aida-I stx2eastA paa
	42	estB aida-I
24	6	estB aida-I astA
	14	estB aida-I astA
51	12	fedA aida-I
	16	estB aida-I astA
255	20	estB aida-I astA
	36	estB aida-I astA paa
17	8	estB aida-I astA
32	24	paa
119	21	paa sepA
284	9	estB
221	3	estB fedA aida-I

^aThe genotypes were determined by REP-PCR, and genotype designations are consistent with those used by Duriez and Topp (16).

^bND, no virulence genes were detected.

E. coli enumeration, isolation, and confirmation.

Isolation of E. coli was performed as previously described (33). Briefly, samples were serially diluted in sterile sodium metaphosphate buffer and plated on mFC-BCIG agar, made with mFC basal agar (Difco, Fisher Scientific, Ottawa, ON, Canada) and 100 μg of BCIG (5-bromo-4-chloro-3-indolyl-β-d-glucuronide cyclohexyl ammonium salt; Medox Diagnostics, Ottawa, ON, Canada) per liter; samples were incubated overnight at 44.5°C, isolated a first time on the same medium, and then restreaked twice on LB agar (Difco, Fisher Scientific, Ottawa, ON, Canada). Isolates were considered to be E. coli if they grew at 44.5°C, had a positive reaction for β-glucuronidase (blue colonies on mFC-BCIG agar), fermented lactose, and produced indole. Isolates confirmed to be E. coli were stored in 15% glycerol at −70°C. Approximately 100 isolates were picked from each sample when enough bacteria were available on the plates.

Molecular methods.

Cell suspensions of E. coli were used as templates in the PCR as follows. Portions of fresh LB broth (100 μl per well) in a sterile 96-well microtiter plate were inoculated from stock cultures. Cells were grown statically at 37°C overnight and centrifuged at 710 × g for 25 min (Centra CL3 microplate centrifuge; Thermo IEC, Needham Heights, MA). The cells were resuspended in 100 μl of sterile Milli-Q H₂O and agitated at 1,000 rpm with a microplate shaker (Sarstedt, Montréal, QC, Canada) for 5 min. The resuspended cells were used directly as a template for the PCR or frozen at −20°C until required.

Virulence genes were detected by multiplex PCR using primers described by Boerlin et al. (6). The 12 pairs of primers were arranged in three groups that each yielded a mix of PCR products varying sufficiently in size to be readily resolved in agarose gels. The first targeted the genes estB, estA, elt, and faeG. The second targeted fanA, fedA, aida-I, and stx_2e. And the third targeted astA, paa, fasA, and sepA. The PCRs were carried out in 1× PCR buffer with (NH₄)₂SO₄ (Fermentas, Burlington, ON, Canada), 3 mM MgCl₂, a 200 μM concentration of each deoxynucleoside triphosphate (Invitrogen, Burlington, ON, Canada), a 2 μM concentration of each primer, 2.5 U of Taq DNA polymerase (Sigma-Aldrich Canada Ltd.), and 2 μl of E. coli cell suspension as a template. Amplifications were performed with a Thermo MBS Satellite 0.2 Thermocycler instrument (VWR International, Mississauga, ON, Canada) as follows: after an initial denaturation at 95°C for 15 min, 30 cycles of denaturation (95°C for 1 min), annealing (at 55°C for 1 min plus 3 s per cycle for multiplex groups 1 and 2; at 62°C 90 s for multiplex group 3), and extension (72°C for 2 min) were performed, followed by a final extension (72°C for 10 min). Every PCR experiment included one negative control and one positive control for each gene as described in Boerlin et al. (6).

REP-PCR fingerprinting was done with the BOXA1R primer as described by Versalovic et al. (53). The final reaction mixture (25 μl) consisted of 1× PCR buffer (Promega, Madison, WI), 1.5 mM MgCl₂, 1% dimethyl sulfoxide, a 200 μM concentration of each deoxynucleoside triphosphate (Invitrogen, Burlington, ON, Canada), 2 μM BOXA1R primer, 1 U of Taq polymerase (Promega), and 2 μl of E. coli suspended cells as a template. Amplification was performed as follows: after an initial denaturation at 94°C for 10 min, 34 cycles of denaturation (at 94°C for 3 sand at 92°C for 30 s), annealing (50°C for 1 min), and extension (65°C for 8 min) were performed, followed by a final extension (at 65°C for 8 min). Six microliters of loading dye was added to 25 μl of PCR product, and 7 μl of this mixture was loaded into wells prepared with an 8-mm by 1-mm comb tooth size. Every eighth well received the MassRuler DNA ladder (Fermentas, Burlington, ON, Canada). PCR products were resolved by horizontal gel electrophoresis (2.5 V/cm for 16 h) in 1.5% agarose gels in 1× Tris-borate-EDTA buffer. The gel was stained with 1 μg ml⁻¹ ethidium bromide solution for 10 min and destained in Milli-Q water for 10 min. Gel images were captured as 16-bit TIFF images, using Alphaease FC software and an Alpha Innotec digital gel documentation system (Fisher Scientific, Ottawa, ON, Canada).

Computer-assisted image and data analysis.

Normalization of gel images and assignment of fingerprints to isolates were done with the Bionumerics software package (version 4.5; Applied Maths, Kortrijk, Belgium) (16). Filtering and background subtraction were optimized for each image independently according to methodology available at http://www.ecolirep.umn.edu/addinggelimages.shtml. The positions of fingerprints on gels were normalized using the MassRuler DNA ladder as the external standard in the range of 400 bp to 4,000 bp. The assignment of strains to different clusters was performed by calculating the similarity coefficients with the curve-based Pearson similarity coefficient. Similarity trees were generated using the unweighted-pair group method using average linkage. Groups with 80% similarity were then created, and final assignments were made on the basis of careful eye examination and analysis of the one-to-one similarity between isolates. Clusters were assigned based on formerly published results (16), and new clusters were added when necessary.

All data were grouped in an Excel database to perform basic statistical analyses. Odds ratios and statistical tests were calculated using Statsdirect, version 2.6.3 (StatsDirect Ltd., Altrincham, United Kingdom).

RESULTS

Frequency and distribution of virulence genes in E. coli populations in freshly shed feces and in stored manure.

E. coli isolates were collected from freshly shed feces (n = 2,193) and from stored manure (n = 2,475) obtained from a single commercial farm over a 6-month period (16). Eight (estB, estA, fedA, stx_2e, astA, paa, aida-I, and sepA) of the 12 (elt, fasA, faeG, and fanA were not detected) virulence genes evaluated by PCR were present in a random sampling (fresh feces, n = 412; stored manure, n = 392) of the E. coli collection. With the exception of estA, the virulence genes were carried at significantly higher frequencies in the isolates from the fresh feces than in the isolates from the stored manure (Table 2). Within the 11 major REP-PCR-defined genotypes, the average number of virulence genes detected per E. coli isolate was always greater in isolates from fresh feces (1.49) than in isolates from stored manure (0.36). Twice as many fecal isolates (n = 223; 54% of the population) carried one or more virulence genes than isolates obtained from the stored manure (n = 109; 27% of the population) (Chi-square test, P < 0.001).

TABLE 2.

Frequency of virulence gene carriage in E. coli isolated from fresh feces and from the manure holding tank of a commercial swine farm

Isolate source (n)a	Frequency (%) of carriageb
	estB*	estA	fedA*	stx2e*	astA*	paa*	aida-I*	sepA*
Fresh feces (412)	28	2	34	8	23	23	20	7
Stored manure (392)	0.5	1	6	0.5	17	7	3	1

^an, number of isolates.

^bNone of the isolates carried elt, fasA, faeG, or fanA. *, frequencies of carriage are significantly different (Fisher's exact test, P < 5%).

Virulence genes were dispersed widely in many distinct E. coli genotypes defined by REP-PCR (16; also data not shown). For example, among the 68 isolates of genotype 16 that were evaluated, nine different virulence gene profiles were detected (no virulence genes, astA, estB aida-I, estB astA aida-I, estB fedA astA aida-I, estB fedA astA paa aida-I, estB fedA paa, fedA paa sepA, and paa sepA). Likewise, in genotype 10, 7 different profiles (no virulence genes, astA, estA, estB, fedA, estB fedA paa, and fedA aida-I) were detected in the 51 isolates evaluated. The gene astA alone was detected in 62 isolates representing 28 distinct genotypes.

Changes in E. coli population structure and loss of virulence genes during storage of fresh fecal material.

The viable E. coli populations in the slurry numbered (5.4 ± 1.6) × 10⁶ bacteria ml⁻¹ at the beginning of the experiment and declined to (1 ± 0.7) × 10⁵ bacteria ml⁻¹ after 7 weeks of incubation in the isolation chambers when the manure holding tank was emptied and the chambers could not be incubated any longer. At the start of the incubation the population was largely dominated by a single genotype, designated 276 (representing 78% of the 181 isolates taken from the freshly prepared slurry) (Table 3). After 1 week in the manure holding tank, the population became largely dominated by genotype 225 (70% and 78% of the 91 isolates obtained from each of the two chambers), which had initially represented only 1% of the population. The dominance of genotype 225 carried through to the end of the incubation (80% and 83% of the isolates in the two chambers after 7 weeks). Genotype 276 declined in significance in the first week (8% and 3% of isolates from the two chambers) and was undetected after 7 weeks. The transition between genotypes 276 and 225 is not the consequence of the loss of the studied virulence genes, as multiple passages of isolates belonging to genotype 225 led to the loss of those virulence genes but not to any detectable changes in the fingerprint (data not shown).

TABLE 3.

Distribution of virulence genes among the major genotypes during the in situ incubation of fresh fecal slurry in a commercial manure holding tank and in parallel laboratory microcosm incubations

Genotype	Virulence genotypea	Frequency (%) of carriage at the indicated wkb
		Wk 0 (n = 181)	Isolation chamber		Laboratory microcosm
			Wk 1 (n = 182)	Wk 7 (n = 105)	Wk 1 (n = 256)	Wk 7 (n = 214)
225	aida-I				0.5 ± 0.8	0.4 ± 0.7
	aida-I astA		0.5 ± 0.8
	astA		11.0 ± 0.0	5.0 ± 7.1		5.9 ± 2.4
	estB aida-I astA	1.2			0.9 ± 1.6
	estB fedA aida-I				0.5 ± 0.8
	fedA aida-I		3.8 ± 3.9		2.5 ± 1.6
	fedA aida-I astA	1.2	1.1 ± 1.6
	ND		57.7 ± 0.8	76.3 ± 9.0	11.2 ± 12.0	54.0 ± 21.1
276	estB fedA aida-I astA				0.5 ± 0.8
	fedA aida-I	77.9	5.5 ± 3.1		46.9 ± 36.7
	ND	0.6			0.5 ± 0.8
221	fedA aida-I				4.1 ± 4.0
	fedA aida-I astA	1.2
	ND		10.4 ± 0.8		8.1 ± 10.7
33	astA					0.9 ± 1.6
	ND					7.0 ± 12.2
285	estB fedA aida-I astA				0.5 ± 0.8
	fedA aida-I				0.5 ± 0.8
	ND	1.7		3.3 ± 4.7	1.7 ± 1.5	3.3 ± 5.8
51	fedA aida-I	1.7
	ND	4.7	1.1 ± 1.6			4.4 ± 4.2
224	fedA aida-I	4.7			0.4 ± 0.6
	ND	1.2	1.6 ± 0.8		0.4 ± 0.6
16	fedA aida-I	1.2			0.9 ± 1.6
	ND		0.5 ± 0.8		1.4 ± 2.4	1.5 ± 1.7
3	ND		0.5 ± 0.8	7.7 ± 3.3	0.5 ± 0.8	0.4 ± 0.7
248	aida-I					2.2 ± 3.8
	ND					2.6 ± 2.3

^aND, no virulence genes were detected.

^bResults represent data from two isolation chambers and three laboratory microcosm experiments. Data are from at least 1% of the total number of isolates of the experiment. Week 0 represents the initial inoculum. Values for weeks 1 and 7 are the averages ± standard deviations. Where no values are given, the frequency of this combination of virulence gene(s) and genotype was below the detection level. n, number of isolates.

At the start of the incubation, 98% of the isolates (n = 181) carried both the aida-I and the fedA genes, including both isolates of genotype 225. Following a week's incubation, the frequency of isolates that carried aida-I and fedA dropped to 13% and 10% (chambers 1 and 2, respectively) and to below 1% in both chambers by week 7. This decrease in frequency was largely associated with the replacement of the dominant genotype 276 (that carried these two genes at high frequency) with genotype 225. Seven percent of recovered genotype 225 isolates carried fedA and aida-I after a week's incubation, and none carried these genes at the end of the experiment. All isolates of genotype 276 recovered throughout the experiment carried fedA and aida-I.

In a parallel incubation, triplicate portions of the freshly prepared slurry were incubated in the laboratory. The viable population of E. coli declined from (5.4 ± 1.6) × 10⁶ at the start of the incubation to (4.2 ± 2.8) × 10³ by week 7. In agreement with the in situ manure storage tank incubation, the dominant genotype 276 at the start of the incubation was rapidly replaced by genotype 225 (Table 3). The shift was accompanied by the decline in the frequency of carriage of fedA and aida-I from 98% of isolates at the start of the incubation to 67% ± 27% at week 1 and then to an undetectable level by week 7.

Changes in population structure and carriage of virulence genes during in situ incubation of a reconstituted E. coli community in a manure storage holding tank.

A reconstituted E. coli community was created by inoculating sterile manure slurry with an equal mixture of 48 E. coli isolates. The inoculated slurry was added into two isolation chambers that were immersed in the manure holding tank of the farm. At the beginning of the experiment the viable population was (1.1 ± 0.1) ×10⁷ bacteria ml⁻¹, and after 3 weeks in the manure holding tank the population had decreased to (1.1 ± 0.5) × 10⁵ bacteria ml⁻¹. The incubation was interrupted as the population represented less than 1% of the initial inoculum.

The two chambers initially contained a diversity of genotypes (Table 4). During the incubation the communities became dominated by a genotype (designated 41) that had initially represented only 6% of the starting population (3 of 48 isolates). Following 1 week of incubation, genotype 41 represented 48% and 43% of the isolates recovered from the two chambers, and by the third week it represented 79% and 81% of the recovered isolates.

TABLE 4.

Distribution of genotypes and virulence profiles during the incubation of the reconstituted E. coli community in the manure storage tank

Genotype designation	Virulence genotypea	Frequency (%) of detection at:
		Wk 0 (n = 48)b	Wk 1 (n = 150)c	Wk 3 (n = 142)c
41	ND		17	79
	estB aida-I astA		26	1
	astA		1
	estB aida-I	2
	astA paa		1
	estB aida-I paa		1
	estB fedA aida-I astA paa		1
	estB fedA aida-I stx2eastA paa	2
	fedA aida-I	2
24	ND		24	11
	estB aida-I astA	4	1
	astA		1
283	fedA stx2eastA paa	6	12	2
	estB fedA aida-I stx2eastA paa	15
	ND	4		1
	fedA stx2e		1
	estB fedA aida-I astA paa	2
	estB astA paa		1
282	estB aida-I astA	13
	estB fedA aida-I stx2eastA paa	4
	estB fedA aida-I astA	4
	ND		1
	estB fedA aida-I	2
	estB aida-I paa		1
	estB fedA aida-I astA paa	2
	estB fedA aida-I paa	2
	astA		1
	fedA astA	2
	fedA paa	2
	fedA stx2e	2
51	astA		5	4
	paa		1
	aida-I astA paa		1
	estB aida-I astA	2
	fedA aida-I	2
255	paa		5	1
	astA paa		1
	estB aida-I astA	2
	estB aida-I astA paa	2
90	estB fedA aida-I astA	4
	estB fedA aida-I	4
	ND		1
	estB paa sepA	2
221	estB fedA aida-I	2
	aida-I astA		1
32	paa	2
17	estB aida-I astA	2
284	estB	2
119	paa sepA	2

^aND, no virulence genes were detected.

^bValues are given only for the gene combination present in the initial inoculum. n, number of isolates.

^cWhere no values are given, the frequency of this combination of virulence gene(s) and genotype was below the detection level. n, number of isolates.

The apparent virulence gene carriage in isolates belonging to the same genotype varied during the incubation (Table 4). Most isolates recovered after 3 weeks carried fewer virulence genes than those isolated after 1 week in the manure holding tank. The three isolates from genotype 41 carried two (fedA or estB plus aida-I) or six (estb, fedA, aida-I, stx_2e, astA, and paa) virulence genes at the beginning of the experiment and less than two on average (between zero and five) at the end of the first week; after 3 weeks only 1 of the 114 isolates of this group carried any virulence genes (estB, aida-I, and astA). Likewise, there were two isolates from genotype 24 in the inoculum, both carrying estB, aida-I, and astA. Following 1 week only one of the 38 isolates of genotype 24 still carried these genes, and after 3 weeks none of the 16 genotype 24 isolates (11% of the population) carried any of the virulence genes. In only three instances were virulence genes found in genotypes that did not include those genes in the starting inoculum. Two of the isolates from genotype 51 were paa positive at the first week, and one from genotype 221 was positive for astA in the first week (Table 4).

Thirty-four of the 187 isolates (18%) recovered after 1 week of incubation had REP-PCR fingerprints that were not present in any of the isolates comprising the inoculum, and 25/167 (15%) of isolates obtained following 3 weeks likewise had fingerprints not present in any of the isolates in the inoculum.

DISCUSSION

Based upon evidence obtained using three complementary experimental approaches, the widely employed commercial practice of storing swine manure slurry statically in anoxic lagoons promotes the loss in E. coli populations of a variety of genes associated with the virulence of this bacterium for swine. First, estB, fedA, stx_2e, astA, paa, aida-I, and sepA were each detected significantly more frequently in freshly shed feces obtained over a 6-month period from a commercial swine herd than in the manure storage tank collecting the waste from that same herd. Second, when freshly shed feces were incubated within the storage tank in a diffusion chamber that allowed contact with the surrounding manure only through 0.45-μm-pore-size membranes, the frequency of detection of fedA and aida-I in viable E. coli went from nearly 100% to below the detection limit. The same result was observed in parallel laboratory incubations of the same fecal slurry. Finally, a defined mixture of pure cultures of swine E. coli incubated in diffusion chambers in the manure storage tank became dominated with isolates that lacked virulence genes.

Our results indicate that the decrease in frequency of virulence genes detected in E. coli populations recovered from manure is mainly due to the loss of genes from specific clonal populations. The clearest evidence for this comes from the experiment with the reconstituted E. coli population. The loss of virulence genes observed in both genotypes 41 and 24 (Table 4) over the 3 weeks of the incubation clearly indicates that virulence genes carried by specific genotypes were lost during manure storage. A similar loss of virulence genes among isolates with the same genotype was also observed in the stored slurry, where the frequency of association of these genes with genotypes 225, 221, and 224 decreased over the course of the experiment. The distribution of all the observed virulence genes, with the exception of astA, decreased during manure storage. Although the astA gene was detected at a somewhat lower frequency in stored manure than in freshly shed fecal material, it was detected in the stored manure at the highest frequency of the virulence genes under investigation (Table 2) and increased in frequency during the incubation of the fecal slurry (Table 5). Nevertheless, all the isolates carrying astA in the stored slurry belonged to genotype 225 and represented only 10% to 15% of this group after storage, a proportion that could not have been reliably detected in the limited number of isolates (n = 4) of the same genotype at the beginning of the experiment. Therefore, it cannot be ruled out that astA was already present from the start. Finally, the role of astA in virulence is still somewhat unclear, and it has been detected with relatively high frequency in commensal E. coli (37). Overall, we observed clear evidence for gene loss in E. coli populations during storage and, even considering astA, no convincing evidence for gene acquisition.

TABLE 5.

Frequency of carriage of virulence genes in fresh fecal slurry during storage in a manure holding tank

Gene	Frequency (%) of carriage under the indicated conditionsa
	Wk 0 (n = 184)	Isolation chamber 1		Isolation chamber 2		Microcosmb
		Wk 1 (n = 96)	Wk 7 (n = 93)	Wk 1 (n = 94)	Wk 7 (n = 94)	Wk 1 (n = 94)	Wk 7 (n = 94)
estB	1					5 ± 5
fedA	98	10		13		67 ± 27
aida-I	98	13		13	1	69 ± 26	2 ± 3
stx2e
astA	1	12	11	13		3 ± 6	8 ± 3
paa
fasA
sepA

^aWeek 0 represents the initial inoculum. Where no values are given, the frequency of the gene was below the detection level. n, number of isolates.

^bData from the three laboratory microcosms were pooled and are expressed as the averages ± 1 standard deviation.

In previous work on these same collections, we demonstrated strong temporal trends in the distribution of antibiotic resistance profiles in freshly shed E. coli and significantly more stability in isolates from the manure storage lagoon (16). There was no relationship between the distribution of REP-PCR fingerprints and the distribution of antibiotic resistance profiles, suggesting that specific antibiotic resistance determinants were dynamically distributed within the population (16). The loss of virulence genes evaluated in the study reported here indicates that they are carried on mobile genetic elements. The strong link between fedA and aida-I in the fecal slurry in situ incubations reported here is consistent with their reported colocation alone on a plasmid (35) or in association with stx_2e (40). Preliminary results in our laboratory indicate that in some isolates the loss of fedA and aida-I is associated with the loss of a unique plasmid (data not shown). Our data do not allow us to determine if bacteria that have shed these virulence genes had a fitness advantage over bacteria that had retained them. Attempts to compare the relative persistence of isolates that varied in their complement of virulence genes using in situ incubations with isolation chambers were foiled by the rapid loss of the genes.

We detected evidence of contamination of the diffusion chamber contents from the surrounding manure. In the reconstituted community experiment, this consisted of the detection of small numbers of enterococci and the appearance at the end of the experiment of E. coli genotypes that were not present in the inoculum. It is not known whether the contamination was due to a small breach in the chamber or to movement of small cells through the membranes or whether it occurred during sampling of the chambers. At no time did contaminating E. coli (detected on the basis of REP-PCR) represent more than 18% of the total isolates recovered. The laboratory microcosm experiments with fecal slurry, clearly not subject to contamination, showed very similar results to those observed in the isolation chambers in terms of clonal replacement and virulence gene loss. Overall, although these results cast doubt on the physical integrity of the diffusion chambers, they do not invalidate the fundamental observations concerning the loss of virulence genes during incubation in the holding tank or their distribution within the E. coli populations as distinguished by REP-PCR (Tables 3 and 4).

We detected a loss of virulence genes upon repeated transfer of isolates on laboratory media (data not shown). There is, therefore, a concern that some of the gene loss we report here occurred following isolation rather than in the manure holding tank or laboratory incubations. However, the same isolation and purification procedures were used in all experiments, and therefore the pressure for gene loss on laboratory media was uniform across all experiments. Significant differences between treatments (barn versus holding tank) and temporal trends (gene loss during incubation in the manure holding tank or laboratory microcosm) with respect to virulence gene distribution were therefore independent of any background loss upon cultivation in the laboratory.

There are numerous reports on the distribution of virulence genes in E. coli obtained from pigs (6, 34, 41, 44), humans (7, 13, 15), other animals (12, 47), or the environment (27, 39). But longitudinal studies on the carriage of virulence genes in changing environments are rare. In a survey of the distribution of 20 genes encoding adherence factors, toxins, invasins, capsules, and iron uptake systems in isolates of E. coli, the frequency of detection was generally lower in environmental isolates than in fecal isolates from human volunteers (39).

Although the virulence genes studied here are more frequently associated with swine pathologies, elt, estA, astA, and sepA are also found in E. coli associated with human pathologies (30), or they have close homologs involved in human pathologies. Hence, although the swine isolates may not be of direct relevance to public health, they could act as a reservoir for virulence genes that could be transferred to strains more directly pathogenic for humans. The strong and stable association of genotype 276 and fedA and aida-I was the only observed occurrence of a stable association between virulence genes and a specific genetic background. It has previously been suggested that the stability of newly acquired genes, including virulence genes, requires coadaptation between the genetic background and the new genes (19), and the viability of numerous virulence genes for strain typing is the proof that such associations can be stable. If the results reported here can be generalized, it appears that during an enteric pathogen's journey from the end of one host's digestive tract to the beginning of the next may in some cases be accompanied by genetic loss that could attenuate virulence potential. This process would mitigate the risk to public health, a factor that has not been considered in microbial risk assessments and source water protection strategies (25). Furthermore, an understanding of environmental conditions that promote the loss of virulence genes might inform new risk management strategies to the benefit of public and veterinary health.