Dynamics of Escherichia coli virulence factors in dairy herds and farm environments from a longitudinal study in the United States

Transcription

1 AEM Accepted Manuscript Posted Online 24 April 2015 Appl. Environ. Microbiol. doi: /aem Copyright 2015, American Society for Microbiology. All Rights Reserved. 1 2 Dynamics of Escherichia coli virulence factors in dairy herds and farm environments from a longitudinal study in the United States Elisabetta Lambertini 1,2, Jeffrey S. Karns 3, Jo Ann S. Van Kessel 3, Huilin Cao 1, Ynte H. Schukken 4,5, David R. Wolfgang 6, Julia M. Smith 7, Abani K. Pradhan 1,2 # 1 Department of Nutrition and Food Science, University of Maryland, College Park, Maryland, USA 2 Center for Food Safety and Security Systems, University of Maryland, College Park, Maryland, USA 3 Environmental Microbial and Food Safety Laboratory, ANRI, USDA-ARS, Beltsville, Maryland, USA 4 Department of Population Medicine and Diagnostic Sciences, Cornell University, Ithaca, New York, USA 5 GD Animal Health, Deventer, Netherlands 6 Department of Veterinary and Biomedical Science, Pennsylvania State University, University Park, Pennsylvania, USA 7 Department of Animal Science, University of Vermont, Burlington, Vermont, USA Running title: E. coli virulence factors in dairy farms Keywords: E. coli, virulence factors, dairy, farm, cow, milk, dynamics, reservoir # Address correspondence to: Abani K. Pradhan, Department of Nutrition and Food Science, University of Maryland, 0112 Skinner Building, College Park, MD 20742, USA. akp@umd.edu 1

2 27 ABSTRACT Pathogenic E. coli or its associated virulence factors have been frequently detected in dairy cow manure, milk, and dairy farm environments. However, it is unclear what the long-term dynamics of E. coli virulence factors are, and which farm compartments act as reservoirs. This study assessed occurrence and dynamics of four E. coli virulence factors (eae, stx1, stx2, and γ- tir) in three U.S. dairy farms. Fecal, manure, water, feed, milk, and milk filter samples were collected from 2004 to Virulence factors were measured by post-enrichment quantitative PCR. All factors were detected in most compartments in all farms. Fecal and manure samples showed the highest prevalence, up to 53% for stx and 21% for γ-tir in fecal samples, up to 84% for stx and 44% for γ-tir in manure. Prevalence was low in milk (up to 1.9% for stx, 0.7% for γ- tir). However, 35% of milk filters were positive for stx and 20% for γ-tir. All factors were detected in feed and water. Factor prevalence and levels, expressed as qpcr cycle threshold categories, fluctuated significantly over time, with no clear seasonal signal independent from year-to-year variability. Levels were correlated between fecal and manure samples, and in some cases autocorrelated, but not between manure and milk filters. Shiga-toxins were nearly ubiquitous, and 10-18% of the lactating cows were potential shedders of E. coli O157 at least once during their time in the herds. E. coli virulence factors appear to persist in many areas of the farms, and therefore contribute to transmission dynamics. 2

3 45 INTRODUCTION Diarrheagenic Escherichia coli of several pathovars of public health importance such as enteropathogenic E. coli (EPEC), Shiga toxigenic E. coli (STEC), and enterohaemorrhagic E. coli (EHEC) have been observed in dairy herds (1-3), as well as in milk (4 10) and other dairy products (4-8, 11, 12). Beef cattle are also known to harbor pathogenic E. coli (13 16). Dairy animals also enter the meat production chain, contributing to meat-borne infections (17). Infection by these classes of pathogenic E. coli can have serious health impacts in humans (8). The cost of human health losses in the United States due to E. coli O157 alone was estimated to amount to $405 million per year (18). Numerous E. coli outbreaks have been linked to the consumption of milk and dairy products (19 22), and to direct contact with dairy farm animals and environments (23, 24). Milk contamination is usually due to fecal contamination. Intestinal colonization by STEC such as E. coli O157 is usually subclinical in cows and calves, and therefore is often undetected. STEC and EHEC have almost never been associated with non-enteric infections in cows, such as mastitis, although other classes of pathogenic E. coli have been known to cause mastitis (25). A virulence factor is a phenotypic trait, usually a large molecule or complex, which determines the ability of E. coli and other bacterial pathogens to infect a host. Common phenotypes include the ability to attach to cells of the intestinal lining, the ability to enter such cells, and the ability to cause damage to the cells, e.g. by secreting toxins (26). The genes encoding E. coli virulence factors are located either on plasmids, on large genome regions called pathogenicity islands ( kb), or on integrated bacteriophages (27), all of which may be exchanged via horizontal 3

4 gene transfer. The genes encoding Shiga toxins 1 and 2 (stx1, stx2) are carried by lamboid phages and can be integrated into the bacterial host genome in multiple copies (28). Genetic elements responsible for attachment to and penetration (effacement) into the host intestinal lining cells are located in the chromosomal locus of enterocyte effacement (LEE), a pathogenicity island which includes the Intimin (eae) factor (E. coli attachment and effacement) and the tir factor (translocated intimin receptor), as well as other virulence elements (29). Other genetic elements are also associated with STEC virulence, including hemolysin, type III secretion factors, and catalase (29, 30). The traditional classification of enteropathogenic E. coli is based on infection mechanisms or symptoms (e.g. EHEC, STEC). To some extent the presence of certain virulence factors can be linked to infection mechanisms and hence pathogenic classes (31). For example, the presence of the intimin gene eae, without any Shiga-toxin genes, defines enteropathogenic (EPEC) strains. The presence of stx factors defines the STEC class, of which enterohaemorrhagic E. coli (EHEC) is a subgroup. The γ-alleles of eae and tir and the rfbe O157 gene (involved in o-antigen synthesis) are present in the EHEC serogroup O157. The presence of both stx2 and γ-tir in an isolate is suggestive of one of the most common virulent EHEC serotypes, O157:H7. Translated to an entire sample where more than one strain may occur, the detection of stx2 and γ-tir in similar amounts suggests that serotype O157:H7 may be present. In general, information on virulence factors is a better predictor of infection severity than serotype alone (30). Historically, assay development and monitoring efforts have been devoted to E. coli O157:H7, due to its predominance in severe human infections and ease of isolation in clinical labs; however, the risk due to non-o157 STEC and EHEC serogroups should not be underestimated (30). In 4

5 particular, in recent years non-o157 STEC such as the big 6 (O26, O45, O103, O111, O121, and O145) have also received increased attention due to their involvement in human disease (32) Dairy cattle and farms are known reservoirs of both STEC and EHEC, including E. coli O157:H7 (2, 8, 29). Dairy heifers, cows, and calves have been found positive for E. coli O157 and other pathogenic E. coli serotypes across the world, including in the U.S. (8, 33 38), South America, Europe, Oceania, and Japan (30, 39 41). Post-weaned calves (between two and four months of age) appear to be more frequently colonized by O157, with prevalence declining thereafter (42 44). E. coli can enter a dairy farm environment through new herd members, environmental media such as air, water, and soil, wildlife, or organic materials such as cattle feed and bedding. Once an E. coli strain has entered the herd, it can persist in the animals intestine and be excreted in the environment. However, the dynamics and routes of introduction, colonization, and persistence in both animals and the farm environment are not well characterized. The dynamics and routes of spread of genetic elements associated with STEC and EHEC virulence in dairy herds and farm environments are also poorly understood. The occurrence of a variety of E. coli virulence factors has been documented (8). There is evidence of horizontal transfer of genetic elements carrying virulence traits (29, 45), as well as antibiotic resistance traits (46 49), although such transfer has not been observed directly in field conditions. Correlative and mechanistic models have been developed to describe pathogen movement in the environment, such as overland waterborne transport (50, 51), or airborne spread (52). However, no current model describes the spread of zoonotic pathogens within farm 5

6 environments, and no model considers the evolution and exchange of virulence traits. The major obstacle hindering the development of such predictive tools is the dearth of data that can support the quantitative risk assessment of virulence factor spread, persistence, or evolution in different farm environments. Most studies on E. coli pathogenicity in dairy herds and farms have focused on short-term surveys and on occurrence only. The present study provides an unprecedented long-term time series of E. coli virulence factor occurrence and relative abundance in dairy farm compartments. While this type of studies are very time- and resource-intensive and therefore rare, they provide crucial data to build and validate predictive models, provide a long-term contamination baseline in farms managed according to current best practices, and allow to test for long-term temporal patterns that result from farm management practices and environmental conditions. The present study is part of a larger collaborative project between USDA ARS and the Regional Dairy Quality Management Alliance (RDQMA), which aimed to elucidate the dynamics of bovine pathogen Mycobacterium avium subsp. paratuberculosis (MAP) and zoonotic pathogens E. coli, Salmonella spp., Listeria monocytogenes, and Campylobacter jejuni in three bovine dairy farms located in the northeast U.S. (53). The goal of the study presented here was to assess the dynamics of selected E. coli virulence factors (eae, stx1, stx2, γ-tir) in these three dairy herds and farm environments. Specific objectives were to: 1) compare presence and levels of E. coli virulence factors in dairy cow feces, farm environmental compartments (e.g. water, feed, manure accumulated on floors), and bulk tank milk in the three study farms, and 2) assess time trends of the four virulence factors in the herds and the farm environments. 6

7 MATERIALS AND METHODS Study sites. The study was conducted in three dairy farms located in the northeast region of the U.S. Farm A was located in New York State, Farm B in Pennsylvania, and Farm C in Vermont. All three farms participated in the state dairy quality assurance program, were DHIA members with monthly milk testing, implemented Farm Animal Identification and Records, and maintained computerized on-farm disease records. The herds consisted of approximately 330, 110, and 140 adult dairy cows on farms A, B, and C respectively. Farms A and C raised heifers on-site. On Farm B, weaned calves were sent to an off-site heifer grower and then brought back into the herd prior to giving birth to their first calf and commencing milk production. Sampling timeline. Sampling was carried out from January 2004 to December 2011 on Farm A, from March 2004 to August 2012 on Farm B, and from November 2004 to September 2010 on Farm C, as previously described (53). Samples from the main farm compartments (fecal samples from individual cows; composite manure from pens housing milking cows, non-lactating cows, periparturient cows, and calves as well as the milking alleyways; feed; source water and water from drinking troughs) were collected every 6 months in Farms A and C, and quarterly in Farm B. As much as possible, the same locations were sampled over time. A smaller number of additional samples were collected or sent by the farm manager between the main sample collection sessions, e.g. when a cow was joining or leaving the herd. In Farm B, following the detection of Salmonella samples were collected more frequently, approximately every 3-4 months (53). Samples of milk and milk filters were collected weekly. 7

8 Sampling and sample processing. Fecal samples and samples from the farm environment were collected and handled as previously described (53). In brief, fecal samples were collected directly from the rectum using gloved hands, and transferred to 50-ml centrifuge vials. Feed samples ( g) were collected in 2-quart zippered bags. Composite manure was collected from the floor and stored in 50-ml centrifuge vials. Water samples were collected in 500-ml or 1-L bottles. Samples were transported to the laboratory overnight on ice, stored at 4 C, and processed within 24 hrs. For fecal and composite manure samples, 25 g of the sample were transferred to a filtered stomacher bag with 50 ml of buffered peptone water (BPW), pummeled for 2 min in a bag mixer (stomacher). For feed samples, 25 g were placed in filtered stomacher bag with a 1:2 (wt/wt) amount of BPW and pummeled for 2 min. For all three sample types, after stomaching 5 ml of the liquid from the filtered side of the bag were added to 5 ml of double-strength EC broth (Difco Laboratories, Detroit, MI). For water samples, 100 ml of each sample were filtered through a 0.45-µm nitrocellulose filter (43 mm diameter). The filter was put in a tube with 10 ml of 1x EC broth. Tubes were vortexed and incubated at 42 C for hours. After incubation the tubes were vortexed, and 1.5 ml aliquot was transferred to a 1.7-ml microcentrifuge tube. Tubes were then centrifuged at 12,000 x g for 2 min, the supernatant was discarded, and each pellet was frozen at -20 C The collection, shipping, and processing of milk and milk filter samples were described previously (9). In brief, 50 ml of bulk tank milk and in-line milk filters were collected aseptically and shipped with cold packs overnight. For enrichment of E. coli, 5 ml of milk were added to 8

9 ml of double-strength EC medium and incubated at 42.5 C for hrs (10). Milk filters were cut in cm 2 pieces, transferred into a large filtered stomacher bag, and mixed 1:1 (wt/wt) with BPW. The bag was stomached for 2 min, briefly massaged by hand to reposition the filter pieces into the liquid, then pummeled again for 2 min. 5 ml of filtrate from the pummeled filter pieces-peptone water mixture was added to 5 ml of double-strength EC broth and incubated at 42.5 C for hrs. For both milk and filters, 1.5 ml of the enrichment was centrifuged (12,000 x g for 2 min), the supernatant removed, and the resulting pellet frozen at -20 C. DNA extraction. For all samples except milk and milk filters, DNA was extracted from the pellets using a MoBio soil kit (MoBio Inc., Carlsbad, CA). The purified nucleic acids were stored at -20 C until PCR analysis. For milk and milk filter samples, DNA was extracted from the pellets using InstaGene Matrix (Bio-Rad Laboratories, Hercules, CA) following the manufacturer s directions. The final DNA preparations (200 µl) were stored at -20 C prior to analysis. qpcr assays. A TaqMan quantitative PCR assay was used to determine the presence and relative abundance of four virulence factors associated with pathogenic E. coli in the nucleic acids extracted from sample enrichments. Specifically, samples were assayed for Shiga toxin genes 1 and 2 (stx1 and stx2), generic intimin (eae), and the γ-allele of the translocated intimin receptor (γ-tir) following the protocol described elsewhere (9, 10). The relative amplicon amount in the sample was expressed as the number of PCR cycles needed for the fluorescence intensity to raise above a specified threshold (Cycle threshold, Ct). qpcr was run on Stratagene Mx4000 and Mx3005P instruments (La Jolla, CA). Baselines and threshold fluorescence levels were set manually for each dye in an assay using fluorescence plots normalized to the ROX 9

10 reference dye (drn). As the assay was semi-quantitative, no calibration curve was used. These methods are described in details in Karns et al. (10). From the beginning of the project through winter 2005 samples were analyzed using individual assays for eae, stx1 and stx2, but only samples with strong eae and stx2 signals were tested for γ-tir. Only a relatively small percent of the total number of samples was processed this way (28 of 5757 samples assayed in Farm A, 571 of 4736 in Farm B, and 336 of 2834 in Farm C). Such samples were included in the statistical analyses by virulence factor for eae, stx1, and stx2. For the attribution to a pathogenic E. coli category, these samples were included only when negative for all three tested virulence factors (attributed to category non-lee non-stx E. coli NLNS), or when attributed to the lowpathogenicity LEE non-stx E. coli LNS category (see category definitions in the following section). After early 2005, all samples were processed in two multiplex PCR assays as described in Van Kessel et al. (9). One assay measured stx1 and an internal amplification control, and the second measured stx2, γ-tir, and eae. Statistical analyses Relationship between cycle threshold Ct and E. coli pathogenic class. Whole-sample E. coli communities were classified into potential E. coli pathogenicity classes as a function of the relative abundance of the four considered virulence factors. Pathogenicity is here defined as enteropathogenicity only. Samples were classified as follows: (1) NLNS, non-lee non-stx E. coli: none of the four virulence factors was detected; (2) LNS, LEE non-stx E. coli: eae and/or γ-tir was detected, but the sample did not meet the criteria to be classified in the STEC, p-ehec, or p-o157 categories (below); (3) STEC: either stx1 or stx2 were detected, but the sample did not 10

11 meet the criteria for EHEC or O157 (below); (4) p-ehec, potential enterohaemorrhagic E. coli: eae and stx1, or eae and stx2, were present in similar proportions (+/- 3 Ct); (5) p-o157, potential E. coli O157: eae was detected, and stx2 and γ-tir were present in similar proportions (+/- 2 Ct). For this analysis, post-enrichment qpcr cycle thresholds (Cts) higher than 35 were considered non-detects. Analysis by sample type. The following sample sources were compared and further analyzed: fecal samples from individual cows, composite manure samples from pen floors, cattle drinking water from troughs, cattle feed, milk, and milk filters. Post-enrichment qpcr Cts higher than 35 were considered non-detects. Prevalence across different groups was compared using the Fisher exact test, and the difference was considered significant if p-value <0.05. Ct values for all samples that underwent PCR analysis were classified in the following categories: Ct <15, Ct 15-20, Ct 20-25, Ct 25-30, Ct 30-35, and zero (no PCR signal or Ct>35) (upper value included in each category, lower value not included). The category Ct was kept in Fig. 3 and Fig. S2 for visualization purposes. Seasonality. Only samples collected during the main sampling sessions were included in the analysis. The effect of season and year on presence and levels of the considered virulence factors were assessed with a logistic regression model with prevalence or Ct category as the dependent variable, season and year as fixed effects, and an interaction effect between season and year. Seasons were defined as winter (Jan-Feb-Mar), spring (Apr, May, Jun), summer (July, Aug, Sep), and fall (Oct, Nov, Dec). Season or year effects were deemed significant when the p- value was <

12 Trends over time and autocorrelation. For each virulence factor, autocorrelation of prevalence within a sample type over time, i.e. over sampling sessions, was tested using Spearman correlation with lags from -4 to +4 sampling sessions. It was assumed that sampling sessions were uniformly spaced over time. This is approximately true for Farm A (sampled twice per year on April 12 th -May 1 st and October 3 rd -15 th from 2004 to 2010, for a total of 14 sessions) and C (sampled twice per year on April 18 th -May 21 st and November 2 nd -15 th in , 12 sessions), less so for Farm B (sampled in March, June, September, and December in , but not all four times per year in , for a total of 27 sessions). Correlation of prevalence between different sample types over time (fecal vs. milk filters, and manure composite vs. milk filters) was tested using logistic regression. For milk filters, prevalence was calculated over the 3-5 weekly samples collected in the month before a sampling session. Prevalence in milk was often zero, and therefore was not analyzed here. This test was carried out only on fecal, composite, and milk filter samples, due to the low number of samples collected in each sampling session for other matrices. Correlation or autocorrelation was considered notable if the Spearman coefficient was higher than 0.4. All data analyses were carried out using the R software version (54). RESULTS 253 E. coli virulence factor prevalence by farm and sample type All four virulence factors were ubiquitous on the three farms. The overall prevalence significantly differed by farm and sample type (Table 1). The generic eae gene was frequently 12

13 detected in all three farms (58-70% of all samples) while as expected the γ-tir allele was the least frequent (13-21% in all samples). Prevalence of the stx genes ranged from 28% to 43% overall. For all farms and factors, composite manure samples collected from floors and farm surfaces consistently showed a significantly higher prevalence than fecal samples from individual cows (15-37% higher), with the highest difference observed for stx1. Composite samples were the most uniform in prevalence across farms. Feed and trough water were less likely to harbor virulence factors than fecal and composite samples, although the fecal-water difference in prevalence was not always significant. In feed samples, a considerably higher prevalence of stx and γ-tir was observed in Farm B than in the other two farms, a trend that was not observed in any other compartment. Prevalence of all virulence factors was drastically lower in milk than any other sample type, in all farms. However, prevalence in milk filters was in most cases significantly higher than in milk, and at similar levels as fecal samples for γ-tir. Prevalence of all factors in milk filters was also notably higher in Farms A and C than Farm B. It has to be noted that milk and milk filter samples were collected weekly throughout the study, while other sample types were collected only 2-4 times a year. Hence, available data for these two matrices may represent different occurrence and transmission patterns compared to other sample types. The distribution of Ct levels in positive samples showed a high degree of variability within each sample type, as expressed by the range of qpcr Cts (Fig. 1 for stx2 and γ-tir; Fig. S1 in the Supplemental Material for eae and stx1). Samples were enriched prior to qpcr analysis, thus Ct values are only indicative of relative abundance, not absolute concentrations in the original 13

14 samples. In all farms, a small number of fecal samples had Ct values below 20, corresponding to a very high post-enrichment concentration, possibly due to super-shedding events. Positive composite manure samples presented Ct ranges similar to fecal samples, and a slightly lower variability, possibly as a result of dilution and die-off. Other sample types had generally lower maximum levels (i.e. higher Ct values), although rare high levels were also observed in other compartments. Positive trough water samples, although infrequent, presented Ct levels comparable to fecal and composite positive samples, and slightly higher levels of eae. In all farms, for stx2 and γ-tir (less so for eae and stx1) positive milk samples had a much lower variability in Ct values than other compartments, in part likely due to the low number of positive samples. The Ct distribution in positive milk filter samples was notably different from milk samples in both range and shape. Occurrence in different farm locations and age groups Composite, feed, and water samples were collected from different locations on the farms. For composite manure, which was mostly collected from floors, the prevalence of γ-tir was distributed differently across farm locations, for the three farms (Table S1). In Farm A, γ-tir prevalence was in the 25-37% range, and not significantly different across barns that housed cows of different ages (calves, heifers, adult cows, and dry cows). In Farms B and C, however, γ- tir prevalence was significantly higher in calves than in adult cows (61% vs. 38% in Farm B, 51% vs. 37% in Farm C). Heifer manure showed a similar prevalence to manure from calves, and was significantly higher than in adult cow manure (results presented only for Farms A and C, where heifers were housed on-site). Manure collected from the manure pit showed a γ-tir prevalence 14

15 comparable to other locations (31% in Farm A, 51% in Farm C), while in Farm B it was notably lower (28%) than in manure collected in pens housing either calves or adult cows. Comparable trends were observed for stx1 and stx2 in composite manure, where prevalence was considerably higher in manure from calves and heifers than in manure from adult cows in all farms. Feed samples included ingredients, formulated adult feed (total mixed ration, TMR), haylage or silage, and feed for calves, although not all categories were present on all farms. TMR and calf feed were collected at the feed bunk and hence could have been in contact with animals and manure. γ-tir was detected in feed for both calves (7-8%) and adult cows (4-10%). γ-tir was also occasionally detected in some feed components such as haylage (0-10%), corn (7% in Farm C), and other feed ingredients (10% in Farm B). stx1 and stx2 were generally found at higher prevalence in ingredients (e.g. corn, silage, haylage) than in finished feed such as TMR sampled in the pens, on Farms B and C. eae was detected in most feed categories in all farms. Overall only a moderate number of feed samples were collected (121, 279, and 106 in Farms A, B, C respectively) and hence prevalence estimates in feed subcategories should be considered preliminary. Virulence factors were also unevenly distributed in water samples (source drinking water, trough water, on-farm streams). eae was detected in virtually all water categories on all three farms. stx1 and stx2 were detected at similar frequency in water from calf and adult cow troughs (21-30% of all trough water samples were positive for stx1, and 13-22% for stx2), and at a lower frequency in source drinking water from local wells, collected at taps. γ-tir was detected 15

16 in 0-5% of source water samples (0 of 31 samples in Farm A, 1 of 34 in Farm B, and 1 in 21 in Farm C), and at 0-9% in trough water (for trough water prevalence see Table 1). In positive trough water samples, Ct levels were generally comparable with levels in manure positive samples (Fig.1). In Farm B, stx1, stx2, and γ-tir were also detected in standing water, and in a nearby stream. The number of water samples collected was 146, 547, and 108 in Farms A, B, and C respectively, 50-79% being trough water samples. Classification into potential E. coli pathogenic classes The occurrence of E. coli potential pathogenic classes in enrichments, according to the operational definitions adopted in this study, varied by farm and sample type, as well as over time (Fig. 2, Table S2). On all three farms 23-29% of all samples did not contain any of the considered virulence factors, and were classified as containing non-lee non-stx E. coli (NLNS). For fecal samples the percentage of NLNS samples was much lower in Farms B and C ( %) than in Farm A (21.0%). All three farms had a sizeable portion of LNS samples (20-31%), i.e. which contained eae and/or γ-tir, but no stx, and could not be classified in any higher pathogenicity class. For fecal samples, Farm A had a higher proportion of LNS, followed by lower proportions of STEC (occurrence of stx1 and/or stx2) and p-ehec (occurrence of eae and stx1 and/or stx2 at approximately the same Ct levels). Conversely, Farm B had the highest proportion of STEC samples among the three farms (40.4%). Farm C had comparable prevalence of LNS and STEC, and the highest prevalence of p-ehec in the three farms (27.7%). In all farms only a minor proportion of samples could be classified as potentially containing E. coli O157 (p- O157, defined by the occurrence of stx2 and γ-tir in similar abundance, with eae also present in 16

17 the enrichment), with the vast majority being from manure composite samples. p-ehec and p- O157 were notably more abundant in manure composite than fecal samples. Conversely, trough water had a very low proportion of potentially pathogenic E. coli, with no p-o157 except for two samples in Farm B and one in Farm C. E. coli in feed followed a different distribution from other sample types, with sample proportion declining from lower-pathogenicity LNS to p- EHEC and p-o157 in all farms. Milk and milk filter samples showed a relative trend similar to what observed between fecal and composite manure samples (reflecting the fact that milk filters represent a composite sample of the milk processed during a milking session): the distribution of E. coli classes was highly skewed towards NLNS in milk (85-95%), while milk filters showed a higher proportion of samples of higher potential pathogenicity. Seasonality For all three farms and all virulence factors, both season and year effects accounted for a significant amount of the variability in prevalence and Ct levels. However, no individual year or season was consistently higher or lower for all farms, virulence factors, or sample types. For fecal samples the interaction between year and season was statistically significant for all farms and virulence factors, making any potential seasonal effect more difficult to parse out. Logistic regression performed on prevalence or on Ct categories yielded substantially the same results, and hence the term levels is used here. Due to the different sampling frequency in the three farms, fecal samples for all four seasons were available only for Farm B (527, 849, 619, and 852 fecal samples respectively for winter, spring, summer, and fall). On Farm A samples were collected only in winter, spring, and fall (308, 1912, and 2318 fecal samples respectively), while 17

18 on Farm C samples were collected in spring and fall (886 and 829 fecal samples respectively). Seasonal trends were not consistent across farms or virulence factors. For eae and γ-tir, levels were significantly higher in spring than winter in Farms A and B, and higher in fall than spring in Farm C. For stx1, levels in fall were higher than in winter in Farms A and B, and levels in fall were higher than in spring in Farm C. However, levels in spring were lower than in winter in Farm B. For stx2, higher levels were observed in the warmer seasons of spring and summer than in winter in Farm B. However, contrary to what observed for stx1, stx2 levels were lower in fall than in winter in Farm A, and lower in fall than spring in Farm C. A concordant pattern between stx1 and stx2 was observed only in Farm B. No consistent trend was observed between overall warmer (summer and fall) and colder seasons (winter, and possibly spring), although seven of the significant differences were compatible with, and three contrary to, the hypothesis that warmer seasons are associated with higher prevalence or levels of pathogenic E. coli. The same seasonality analysis applied to composite manure samples yielded different results than for fecal samples. Namely, no significant season effect was observed for any of the factors in Farms A and B, and only a few years had significantly higher or lower levels than others. In Farm C, however, levels of all virulence factors were significantly higher in fall than in spring. Time trends and autocorrelation The trend of Ct categories in fecal samples over time-ordered sampling sessions is shown in Fig. 3 for stx2 and γ-tir, and in Fig. S2 in the Supplemental Material for eae and stx1. In all three farms, Ct levels fluctuated visibly over time. Ct levels appear somewhat clustered in time at a 18

19 visual inspection, a trend that is more visible for stx1 and stx2 in all farms. No significant correlation was observed between prevalence in fecal and milk filter samples, or between composite manure and milk filter samples, for any of the factors. No consistent trend in autocorrelation of prevalence was observed across farms or factors. A few instances of positive autocorrelation of Ct means at lag 1 (i.e. contiguous sampling sessions) were observed in fecal samples for eae (Farm B), stx1 (Farm C), and stx2 (Farms B and C), which may indicate sporadic long waves of occurrence at similar levels spanning two sampling sessions. However, in most cases no autocorrelation was observed at lag 1. More instances of positive autocorrelation at lag 1 were observed in manure composite, for both prevalence and Ct means and for eae and stx, but never for γ-tir. Other instances of autocorrelation at different time lags do not correspond to any yearly or seasonal cycles, and are consistent with the high degree of variability over time observed for all factors. DISCUSSION Prevalence of virulence factors and E. coli pathogenic classes To our knowledge, this is the first study to assess the dynamics of E. coli virulence factors on dairy farms over a time frame of several years. This large longitudinal data set provides a longterm perspective on the temporal patterns of prevalence and relative abundance of virulence factors associated with diarrheagenic E. coli, in the major matrices of three dairy farms located in different northeastern U.S. regions. While several past studies assessed the occurrence of E. coli pathovars or virulence factors at a single point in time or over a short period of up to a few 19

20 months (35, 55-60), intensive long-term studies are necessary to understand patterns of a pathogen s introduction, epidemic curve, persistence, and resurgence on a farm. Large longitudinal data sets, in addition to short-term surveys focusing on specific factors, are also crucial for building and validating mathematical models to predict pathogen dynamics and test mitigation strategies. Direct identification of pathogenic E. coli in dairy farm samples through culture of the organism is difficult due to the wide degree of E. coli diversity in manure and feces. Except for O157:H7, isolating STEC strains is difficult due to the lack of metabolic differences that can be exploited for their selection. Even O157 is difficult to isolate from manure and fecal samples, often requiring time-consuming extraction with immuno-magnetic beads and expensive chromogenic agars to increase the probability of successful isolation. Using quantitative PCR to detect four virulence factor genes associated with enteropathogenic E. coli after enrichment provides a direct semi-quantitative comparison of the relative abundance of these virulence factors within the community of E. coli in a sample. Because such comparison describes the community and not individual cells, it cannot predict with certainty the presence of specific pathogenic serotypes, but more generally it describes the population of each virulence factor within the mix of E. coli and other strains that can multiply in enrichments designed for E. coli growth. Therefore, while the presence of a virulence factor combination in a sample does not necessarily mean a specific pathogenic E. coli strain is present, it implies the possibility for such occurrence. The operational definition of E. coli O157 and other E. coli categories used in this study is an attempt to describe this possibility This study confirms the widespread presence of potentially pathogenic E. coli in dairy herds and dairy farm environments. All four E. coli virulence factors considered in this study were 20

21 consistently observed in all sample types and in all three farms, with the exception of γ-tir in milk and milk filters, and most frequently in fecal and manure composite samples. The prevalence of STEC genes (stx1 and stx2), an overall 28-43% in the three study farms, was consistent with and somewhat higher than ranges observed in other STEC virulence factor surveys in U.S. dairy herds (2, 8). However, stx prevalence higher than in this study, up to 58% in manure, was also observed in the U.S. (33). The γ-allele of the tir virulence factor associated with E. coli O157:H7 has also been commonly detected in dairy farms, in the U.S. and other countries (3). A comparison between survey data obtained by culture-based identification methods and PCR targeting virulence factors is not directly appropriate, as significantly different results have been observed (30). In addition, several past studies reported prevalence of E. coli strains in reference to the number of unique isolates, not individual samples. However, under the assumption that the STEC and p-o157 potential pathogenic classes adopted in this study can be taken as a whole-sample proxy for the presence of STEC and O157 strains, the prevalence ranges here reported are consistent with the wide range of STEC prevalence ( %) observed in fecal samples from adult dairy cows in the U.S. (8, 30, 33, 40, 61 63). These figures are also consistent with the range of STEC prevalence in dairy cattle observed in other countries (13, 30, 64, 65). Also E. coli strain O157:H7 has been frequently isolated in dairy herds in the U.S. (34 36, 58) and in other countries (14, 65, 66). It is known that that young cattle harbor STEC at significant prevalence, often higher than adult cows (13, 60, 63, 66-68). Also, one study observed that soil and floor areas where calves were raised had higher prevalence of STEC than areas housing adult cows (69). The higher prevalence of stx1, stx2, and γ-tir observed in calf vs. 21

22 adult cow composite manure in two of the farms appears to confirm this trend. However, in Farm A the considered virulence factors were detected at similar frequencies in composite manure from all age groups In this study stx genes were occasionally detected in bulk tank milk from all three farms and γ- tir was detected in the milk from one farm. Virulence factor patterns compatible with STEC pathovars were detected at low prevalence ( %) in milk, but at higher prevalence in milk filters ( %). STEC strains or stx virulence factors have also been detected at low prevalence in milk from U.S. farms (5, 8, 9, 70) and European farms (20, 71, 72). E. coli O157 or the associated virulence factors have also been detected in raw bulk milk at prevalence in the range of 0.2% to 9.1% (10, 73). Other surveys, however, detected no E. coli O157:H7 in raw milk in the U.S. (6, 35, 70, 74) and abroad (43, 75). In this study, virulence factor patterns compatible with the presence of E. coli O157:H7 were not detected in any of the milk samples, but they were detected in 0-2.0% of milk filters. Overall, existing evidence points to the fact that even when STEC and EHEC are present in dairy cow feces or the farm environment, appropriate sanitary practices can effectively lower the risk of milk contamination. Milk filters sampled in this study showed a much higher prevalence of all four virulence factors than bulk milk, highlighting the impact of different sampling strategies and sample volume on assay sensitivity and observed prevalence, as seen in studies on other pathogens (76 78). Other surveys of milk filters observed a stx prevalence of 13.6% over 15 farms (33), and 51% in a large survey of dairy herds in the U.S. (9). E. coli O157:H7 was not detected by culture in any milk filters sampled by Hancock (35), while virulence factors compatible with the potential presence of E. coli O157:H7 were detected in 4.2% of raw milk samples in the 2002 NAHMS survey (based on the detection 22

23 of eae, stx1 or stx2, and γ-tir, but not using the semi-quantitative criterion adopted in the present study; only 0.1%, i.e. 1 of 859 NAHMS samples was positive for E. coli O157:H7 by a combination of selective culture media and agglutination assays) (10) This study supports a body of evidence showing that E. coli virulence factors associated with STEC, EHEC, and EHEC O157:H7 are commonly present in the dairy farm environment, besides in cow manure. STEC or stx genes have been observed in dairy cow feed in some studies (63, 69, 79), but not in others (37). It is unclear if feed may be a source of pathogenic E. coli contamination, or if it becomes contaminated after being in contact with the animals or other farm environments. STEC has been detected in drinking water used by dairy cows (36, 63, 64, 66, 69, 80 81). E. coli O157:H7 has also been detected in dairy cow feed (79) and trough water (36, 79). The very low frequency of detection of low levels of STEC virulence factors in source water in this study indicates that well water is unlikely to be a vehicle introducing STEC into the farms. Trough water is most likely contaminated by the cows, as shown by the much higher prevalence of STEC virulence factors in trough water than source water across all farms and cow ages. Some studies have found a correlation between trough water contamination by E. coli O157 and infection in cattle herds, but the direction of pathogen spread is unclear (36, 79). Although trough water is very unlikely to be the vehicle introducing pathogens into the herd, E. coli O157:H7 has been shown to survive for weeks in trough water (82), and for several months in simulated trough sediments, even when chlorine was added to the water (83). Hence, water can potentially act as reservoir and as a vehicle spreading pathogens from cow to cow (79). 23

24 Most past surveys assessed prevalence of generic or pathogenic E. coli, but concentrations were determined only in a few studies, for instance in milk (73, 84). One study on fecal shedding by calves observed concentrations of E. coli O157 up to 10 6 CFU/g (85). Since pathogenic E. coli is common on dairy farms, prevalence alone has not proven sufficient to identify critical reservoirs or the likelihood of transmission between reservoirs. Studies directly measuring concentrations or relative abundance, such as the present study, are crucial in both understanding which farm compartments are reservoirs of E. coli pathogenic elements, and in pinpointing when and how frequently spikes in virulence factor abundance occur. Seasonality Temperature-dependent seasonality in the prevalence of pathogenic E. coli including E. coli O157:H7 in dairy herds has been suggested, although only a few studies have investigated time trends in details. Specifically, a higher prevalence of E. coli O157 was observed during warm weather, approximately from May to September in the northern hemisphere (30, 43, 58, 86-91), which is also the time when human outbreaks have occurred at higher frequency (92, 93). Significant seasonal effect on the occurrence of non-o157 STEC or stx genes has also been observed in some studies (37, 94). An increase in pathogenic E. coli prevalence in the warmer months was also documented in feedlot cattle herds (37, 95 99). Based on the analysis of data from this 7-9 year longitudinal study, no consistent seasonality was observed across farms, while a significant interaction effect between seasons and years was present. The autocorrelation analysis also did not provide strong evidence for or against seasonal or yearly cycles. It is possible that the sampling frequency (two sampling sessions per year in Farms A and 24

25 C, and 3-4 sampling sessions per year in Farm B) masked seasonal patterns. However, although the evidence is not conclusive, seven out of ten significant seasonal differences, spanning all four factors, were in favor of the hypothesis that warmer seasons are associated with higher levels of E. coli virulence factors Time trends and transmission dynamics In a short-term study of shedding in naturally-infected calves, intermittent but ongoing shedding of E. coli O157 over a time frame of a few weeks was observed in dairy calf feces, and a small proportion of high-shedding individual calves were observed to shed almost continuously (85). Other studies observed intermittent shedding of STEC by adult dairy cows, each animal potentially shedding for several months (63, ). Similar intermittent shedding patterns have been observed in feedlot cattle (58, 103). Individual and herd shedding patterns have been associated with several factors, including diet, herd size, housing type, and age of the animal (3). While all pathogenic classes were observed consistently over time, in this study prevalence and Ct levels fluctuated widely for all farms and all factors, a pattern that is compatible with intermittent shedding of variable intensity in individual animals, but ongoing at herd level. Occasional time clusters observed in fecal and composite samples are compatible with long shedding peaks spanning consecutive sampling sessions (i.e. time lag 1), although frequent shorter-term peaks cannot be excluded. This pattern is also compatible with the ongoing presence of a factor, with periodic flares in prevalence and levels not clearly associated with seasons. Overall, the long-term nature of the study and the frequency of sampling that was logistically feasible allowed analyzing time dynamics at a maximum time resolution of

26 months. Hence, while the observed long-term dynamics are compatible with underlying shedding patterns identified over the time frames of days and weeks in other studies, no definitive conclusions can be drawn on short-term dynamics occurring between sampling sessions Data on prevalence of virulence factors and E. coli pathogenic classes in different farm compartments and time dynamics can provide insight into prevention and mitigation strategies to reduce the persistence and movement of zoonotic pathogens on dairy farms. For example, the presence of E. coli virulence factors observed in manure, feed, and to a lesser extent water makes these compartments potential candidates for transmission reduction measures. Furthermore, the variability in prevalence observed between different age groups may support physical separation as a measure to reduce transmission. As demonstrated by the low prevalence and low-pathogenicity observed in milk samples in this study, hygienic measures to prevent fecal contamination of milk appear to be effective, although exceptions (and not knowing the cause of such exceptions) can still result in significant risk. While it was outside the scope of this study to test the effect of farm practices, there is evidence that different farm management strategies such as spreading manure on pastures, extent of contact between animals, organic vs. conventional approaches, diet, and hygiene may impact the risk of pathogenic E. coli occurrence and transmission (3, 33, 35, 104, 105). However, such evidence is still ambiguous, and climactic and geographical variables are also potential risk factors (80). While eradication of pathogenic E. coli on dairy farms still appears a far-fetched goal due the high prevalence and widespread geographical distribution of the pathogen, as well as its persistent fluctuating occurrence in several farm compartments, a better understanding of the 26

27 ecology of pathogenic E. coli and the mobile genetic elements associated with its virulence can lead to improved strategies to control E. coli pathogenicity on farms ACKNOWLEDGEMENTS We thank farm owners and personnel that participated in the study. We thank Thomas Jacobs, Jakeitha Sonnier, Crystal Rice-Trujillo, and Claudia Lam for their technical assistance. Downloaded from on June 19, 2018 by guest 27

28 559 REFERENCES Meng J, Zhao S, Doyle MP Virulence genes of Shiga toxin-producing Escherichia coli isolated from food, animals and humans. Int. J. Food Microbiol. 45: Gyles CL Shiga toxin-producing Escherichia coli: an overview. J. Anim. Sci. 85:E Farrokh C, Jordan K, Auvray F, Glass K, Oppegaard H, Raynaud S, Thevenot D, Condron R, De Reu K, Govaris A, Heggum K, Heyndrickx M, Hummerjohann J, Lindsay D, Miszczycha S, Moussiegt S, Verstraete K, Cerf O Review of Shiga-toxin-producing Escherichia coli (STEC) and their significance in dairy production. Int. J. Food Microbiol. 162: Carneiro LAM, Lins MC, Garcia FRA, Silva APS, Mauller PM, Alves GB, Rosa ACP, Andrade JRC, Freitas-Almeida AC, Queiroz MLP Phenotypic and genotypic characterisation of Escherichia coli strains serogrouped as enteropathogenic E. coli (EPEC) isolated from pasteurised milk. Int. J. Food Microbiol. 108: Stephan R, Schumacher S, Corti S, Krause G, Danuser J, Beutin L Prevalence and characteristics of Shiga toxin-producing Escherichia coli in swiss raw milk cheeses collected at producer level. J. Dairy Sci. 91: Jayarao BM, Henning DR Prevalence of foodborne pathogens in bulk tank milk. J. Dairy Sci. 84:

29 Solomakos N, Govaris A, Angelidis AS, Pournaras S, Burriel AR, Kritas SK, Papageorgiou DK Occurrence, virulence genes and antibiotic resistance of Escherichia coli O157 isolated from raw bovine, caprine and ovine milk in Greece. Food Microbiol. 26: Hussein HS, Sakuma T Prevalence of Shiga toxin-producing Escherichia coli in dairy cattle and their products. J. Dairy Sci. 88: Van Kessel JAS, Karns JS, Lombard JE, Kopral CA Prevalence of Salmonella enterica, Listeria monocytogenes, and Escherichia coli virulence factors in bulk tank milk and in-line filters from U.S. dairies. J. Food Prot. 74: Karns JS, Van Kessel JS, McClusky BJ, Perdue ML Incidence of Escherichia coli O157:H7 and E. coli virulence factors in US bulk tank milk as determined by polymerase chain reaction. J. Dairy Sci. 90: Rey J, Sánchez S, Blanco JE, Hermoso de Mendoza J, Hermoso de Mendoza M, García A, Gil C, Tejero N, Rubio R, Alonso JM Prevalence, serotypes and virulence genes of Shiga toxin-producing Escherichia coli isolated from ovine and caprine milk and other dairy products in Spain. Int. J. Food Microbiol. 107: Civera T, Decastelli L, Martorana M, Dalmasso A, Nucera D, Bottero MT Survey on the presence of enterovirulent Escherichia coli in cheese manufactured in the Italian alpine region. Ital. J. Food Sci. 2:

30 Aidar-Ugrinovich L, Blanco J, Blanco M, Blanco J, Leomil L, Dahbi G, Mora A, Onuma D, Silveira W, Pestana de Castro A Serotypes, virulence genes, and intimin types of Shiga toxin-producing Escherichia coli (STEC) and enteropathogenic E. coli (EPEC) isolated from calves in São Paulo, Brazil. Int. J. Food Microbiol. 115: Aslantaş O, Erdoğan S, Cantekin Z, Gülaçti I, Evrendilek GA Isolation and characterization of verocytotoxin-producing Escherichia coli O157 from Turkish cattle. Int. J. Food Microbiol. 106: Derzelle S, Grine A, Madic J, de Garam CP, Vingadassalon N, Dilasser F, Jamet E, Auvray F A quantitative PCR assay for the detection and quantification of Shiga toxinproducing Escherichia coli (STEC) in minced beef and dairy products. Int. J. Food Microbiol. 151: Varela-Hernández JJ, Cabrera-Diaz E, Cardona-López MA, Ibarra-Velázquez LM, Rangel- Villalobos H, Castillo A, Torres-Vitela MR, Ramírez-Alvarez A Isolation and characterization of Shiga toxin-producing Escherichia coli O157:H7 and non-o157 from beef carcasses at a slaughter plant in Mexico. Int. J. Food Microbiol. 113: Troutt HF, Osburn BI Meat from dairy cows: possible microbiological hazards and risks. Rev. Sci. Tech. 16: Frenzen PD, Drake A, Angulo FJ Economic cost of illness due to Escherichia coli O157 infections in the United States. J. Food Prot. 68:

31 Langer AJ, Ayers T, Grass J, Lynch M, Angulo FJ, Mahon BE Nonpasteurized dairy products, disease outbreaks, and state laws - United States, Emerg. Infect. Dis. 18: Claeys WL, Cardoen S, Daube G, Block J De, Dewettinck K, Dierick K, Zutter L De, Huyghebaert A, Imberechts H, Thiange P, De Block J, De Zutter L, Vandenplas Y, Herman L Raw or heated cow milk consumption: review of risks and benefits. Food Control 31: De Buyser M-L, Dufour B, Maire M, Lafarge V Implication of milk and milk products in food-borne diseases in France and in different industrialised countries. Int. J. Food Microbiol. 67: Altekruse SF, Timbo BB, Mowbray JC, Bean NH, Potter ME Cheese-associated outbreaks of human illness in the United States, 1973 to 1992: sanitary manufacturing practices protect consumers. J. Food Prot. 61: Crump JA, Griffin PM, Angulo FJ Bacterial contamination of animal feed and its relationship to human foodborne illness. Clin. Infect. Dis. 35: CDC. Centers for Disease Control and Prevention Outbreak of Shiga toxinproducing Escherichia coli O111 infections associated with a correctional facility dairy - Colorado, Morb. Mortal. Wkly. Rep. 61:

32 Fernandes JBC, Zanardo LG, Galvão NN, Carvalho IA, Nero LA, Moreira MAS Escherichia coli from clinical mastitis: serotypes and virulence factors. J. Vet. Diagn. Invest. 23: Donnenberg M Escherichia coli: virulence mechanisms of a versatile pathogen. Academic Press. San Diego, CA, USA. 27. Hacker J, Kaper JB Pathogenicity islands and the evolution of microbes. Annu. Rev. Microbiol. 54: Bergan J, Dyve Lingelem AB, Simm R, Skotland T, Sandvig K Shiga toxins. Toxicon 60: Caprioli A, Morabito S, Brugère H, Oswald E, Apriolia AC, Orabitoa SM, Rugèreb HB, Swaldb EO Enterohaemorrhagic Escherichia coli : emerging issues on virulence and modes of transmission. Vet. Res. 36: Monaghan A, Byrne B, Fanning S, Sweeney T, McDowell D, Bolton DJ Serotypes and virulence profiles of non-o157 Shiga toxin-producing Escherichia coli isolates from bovine farms. Appl. Environ. Microbiol. 77: Kobayashi N, Lee K-I, Yamazaki A, Saito S, Furukawa I, Kono T, Maeda E, Isobe J, Sugita- Konishi Y, Hara-Kudo Y Virulence gene profiles and population genetic analysis for exploration of pathogenic serogroups of shiga toxin-producing Escherichia coli. J. Clin. Microbiol. 51:

33 Federal Register Food Safety and Inspection Service of the Department of Agriculture. Shiga toxin-producing Escherichia coli in certain raw beef products. Federal Register Vol. 77, No. 105: Cho S, Diez-Gonzalez F, Fossler CP, Wells SJ, Hedberg CW, Kaneene JB, Ruegg PL, Warnick LD, Bender JB Prevalence of Shiga toxin-encoding bacteria and Shiga toxin-producing Escherichia coli isolates from dairy farms and county fairs. Vet. Microbiol. 118: Murinda SE, Nguyen LT, Landers TL, Draughon FA, Mathew AG, Hogan JS, Smith KL, Hancock DD, Oliver SP Comparison of Escherichia coli isolates from humans, food, and farm and companion animals for presence of Shiga toxin-producing E. coli virulence markers. Foodborne Pathog. Dis. 1: Hancock DD, Besser TE, Kinsel ML, Tarr PI, Rice DH, Paros MG The prevalence of Escherichia coli O157:H7 in dairy and beef cattle in Washington State. Epidemiol. Infect. 113: Faith NG, Shere JA, Brosch R, Arnold KW, Ansay SE, Lee MS, Luchansky JB, Kaspar CW, Lee M Prevalence and clonal nature of Escherichia coli O157:H7 on dairy farms in Wisconsin. Appl. Environ. Microbiol. 62: Cobbold RN, Rice DH, Szymanski M, Call DR, Hancock DD Comparison of Shigatoxigenic Escherichia coli prevalences among dairy, feedlot, and cow-calf herds in Washington state. Appl. Environ. Microbiol. 70:

34 Dunn JR, Keen JE, Thompson RA Prevalence of shiga-toxigenic Escherichia coli O157:H7 in adult dairy cattle. J. Am. Vet. Med. Assoc. 7: Ezawa A, Gocho F, Kawata K, Takahashi T, Kikuchi N High prevalence of enterohemorrhagic Escherichia coli (EHEC) O157 from cattle in selected regions of Japan. J. Vet. Med. Sci. 66: Oliveira MG, Brito JRF, Gomes TAT, Guth BEC, Vieira MAM, Naves ZVF, Vaz TMI, Irino K Diversity of virulence profiles of Shiga toxin-producing Escherichia coli serotypes in food-producing animals in Brazil. Int. J. Food Microbiol. 127: EFSA Statistical analysis of temporal and spatial trends of zoonotic agents in animals and food Part I : critical review of the statistical analysis carried out on the Community Summary Report 2006 data. EFSA J. 253: Eriksson E, Aspan A, Gunnarsson A, Vagsholm I Prevalence of verotoxinproducing Escherichia coli (VTEC) O157 in Swedish dairy herds. Epidemiol. Infect. 133: Heuvelink AE, Bleumink B Occurrence and survival of Verocytotoxin-producing Escherichia coli O157 in raw cow s milk in the Netherlands. J. Food Prot. 61: Meyer-Broseta S, Bastian SN, Arné PD, Cerf O, Sanaa M Review of epidemiological surveys on the prevalence of contamination of healthy cattle with Escherichia coli serogroup O157:H7. Int. J. Hyg. Environ. Health 203:

35 Gamage SD, Strasser JE, Chalk CL, Weiss AA Nonpathogenic Escherichia coli can contribute to the production of Shiga toxin. Infect. Immun. 71: Heuer H, Schmitt H, Smalla K Antibiotic resistance gene spread due to manure application on agricultural fields. Curr. Opin. Microbiol. 14: Oppegaard H, Steinum TM, Wasteson Y Horizontal transfer of a multi-drug resistance plasmid between coliform bacteria of human and bovine origin in a farm environment. Appl. Environ. Microbiol. 67: Alcaine SD, Sukhnanand SS, Warnick LD, Su W-L, McGann P, McDonough P, Wiedmann M Ceftiofur-resistant Salmonella strains isolated from dairy farms represent multiple widely distributed subtypes that evolved by independent horizontal gene transfer. Antimicrob. Agents Chemother. 49: Dröge M, Pühler A, Selbitschka W Horizontal gene transfer among bacteria in terrestrial and aquatic habitats as assessed by microcosm and field studies. Biol. Fertil. Soils 29: Tyrrel SF, Quinton JN Overland flow transport of pathogens from agricultural land receiving faecal wastes. J. Appl. Microbiol. 94: Russell Walker FJ, Stedinger JR Fate and transport model of Cryptosporidium. J. Environ. Eng. 125:

36 Ssematimba A, Hagenaars TJ, de Jong MCM Modelling the wind-borne spread of highly pathogenic avian influenza virus between farms. PLoS One 7:e Pradhan AK, Van Kessel JS, Karns JS, Wolfgang DR, Hovingh E, Nelen KA, Smith JM, Whitlock RH, Fyock T, Ladely S, Fedorka-Cray PJ, Schukken YH Dynamics of endemic infectious diseases of animal and human importance on three dairy herds in the northeastern United States. J. Dairy Sci. 92: R Core Team R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria Sasaki Y, Murakami M, Maruyama N, Yamamoto K, Haruna M, Ito K, Yamada Y Comparison of the prevalence of Shiga toxin-producing Escherichia coli strains O157 and O26 between beef and dairy cattle in Japan. J. Vet. Med. Sci. 75: Andrade GI, Coura FM, Santos ELS, Ferreira MG, Galinari GCF, Facury Filho EJ, de Carvalho AU, Lage AP, Heinemann MB Identification of virulence factors by multiplex PCR in Escherichia coli isolated from calves in Minas Gerais, Brazil. Trop. Anim. Health Prod. 44: Irino K, Kato MAMF, Vaz TMI, Ramos II, Souza MAC, Cruz AS, Gomes TAT, Vieira MAM, Guth BEC Serotypes and virulence markers of Shiga toxin-producing Escherichia coli (STEC) isolated from dairy cattle in São Paulo State, Brazil. Vet. Microbiol. 105:

37 Lejeune JT, Besser TE, Rice DH, Berg JL, Stilborn RP, Hancock DD Longitudinal study of fecal shedding of Escherichia coli O157:H7 in feedlot cattle: predominance and persistence of specific clonal types despite massive cattle population turnover. Appl. Environ. Microbiol. 70: Hancock DD, Rice DH, Thomas ANN, Dargatz DA, Besser TE Epidemiology of Escherichia coli O157 in feedlot cattle. J Food Prot. 60: Wells JG, Shipman LD, Greene KD, Sowers EG, Green JH, Cameron DN, Downes, FP, Martin, ML, Griffin, PM, Ostroff, SM, Potter, ME, Tauxe, RV, Wachsmuth, IK Isolation of Escherichia coli serotype O157:H7 and other Shiga-like-toxin-producing E. coli from dairy cattle. J Clin. Microbiol. 29: Fremaux B, Prigent-Combaret C, Vernozy-Rozand C Long-term survival of Shiga toxin-producing Escherichia coli in cattle effluents and environment: an updated review. Vet. Microbiol. 132: Nastasijevic I, Mitrovic R, Buncic S The occurrence of Escherichia coli O157 in/on faeces, carcasses and fresh meats from cattle. Meat Sci. 82: Rahn K, Renwick SA, Johnsom RP, Wilson JB, Clarke RC, Alves D, McEwen S, Lior H, Spika J Persistence of Escherichia coli O157:H7 in dairy cattle and the dairy farm environment. Epidemiol. Infect. 119:

38 Fremaux B, Raynaud S, Beutin L, Rozand CV Dissemination and persistence of Shiga toxin-producing Escherichia coli (STEC) strains on French dairy farms. Vet. Microbiol. 117: Ezawa A, Gocho F, Kawata K, Takahashi T, Kikuchi N High prevalence of enterohemorrhagic Escherichia coli (EHEC) O157 from cattle in selected regions of Japan. J. Vet. Med. Sci. 66: Parma AE, Sanz ME, Blanco JE, Blanco J, Viñas MR, Blanco M, Padola NL, Etcheverría AI Virulence genotypes and serotypes of verotoxigenic Escherichia coli isolated from cattle and foods in Argentina. Importance in public health. Eur. J. Epidemiol. 16: Bibbal D, Loukiadis E, Kerouredan M, Ferre` F, Dilasser F, Peytavin de Garam C, Cartier P, Oswald E, Gay E, Auvray F, Brugere H. Prevalence of carriers of Shiga toxin-producing Escherichia coli serotypes O157:H7, O26:H11, O103:H2, O111:H8 and O145:H28 in French slaughtered adult cattle. Appl. and Environ. Microbiol., in press. 68. Cray WCJ, Moon HW Experimental infection of calves and adult cattle with Escherichia coli O157:H7. Appl. Environ. Microbiol. 61: Polifroni R, Etcheverría AI, Sanz ME, Cepeda RE, Krüger A, Lucchesi PMA, Fernández D, Parma AE, Padola NL Molecular characterization of Shiga toxin-producing Escherichia coli isolated from the environment of a dairy farm. Curr. Microbiol. 65:

39 Jayarao BM, Donaldson SC, Straley BA, Sawant AA, Hegde NV, Brown JL A survey of foodborne pathogens in bulk tank milk and raw milk consumption among farm families in Pennsylvania. J. Dairy Sci. 89: Little CL, Rhoades JR, Sagoo SK, Harris J, Greenwood M, Mithani V, Grant K, McLauchlin J Microbiological quality of retail cheeses made from raw, thermized or pasteurized milk in the UK. Food Microbiol. 25: Pradel N, Bertin Y, Martin C, Livrelli V Molecular analysis of Shiga toxin-producing Escherichia coli strains isolated from hemolytic-uremic syndrome patients and dairy samples in France. Appl. Environ. Microbiol. 74: Jackson EE, Erten ES, Maddi N, Graham TE, Larkin JW, Blodgett RJ, Schlesser JE, Reddy RM Detection and enumeration of four foodborne pathogens in raw commingled silo milk in the United States. J. Food Prot. 75: D Amico D, Groves E, Donnelly C Low incidence of foodborne pathogens of concern in raw milk utilized for farmstead cheese production. J. Food Prot. 71: Coia JE, Johnston Y, Steers NJ, Hanson MF A survey of the prevalence of Escherichia coli O157 in raw meats, raw cow s milk and raw-milk cheeses in south-east Scotland. Int. J. Food Microbiol. 66:

40 Van Kessel JS, Karns JS, Wolfgang DR, Hovingh E, Jayarao BM, Van Tassell CP, Schukken YH Environmental sampling to predict fecal prevalence of Salmonella in an intensively monitored dairy herd. J. Food Prot. 71: Latorre AA, Van Kessel JS, Karns JS, Zurakowski MJ, Pradhan AK, Boor KJ, Jayarao BM, Houser BA, Daugherty CS, Schukken YH Biofilm in milking equipment on a dairy farm as a potential source of bulk tank milk contamination with Listeria monocytogenes. J. Dairy Sci. 93: Van Kessel JAS, Karns JS, Wolfgang DR, Hovingh E Regional distribution of two dairy-associated Salmonella enterica serotypes. Foodborne Pathog. Dis. 10: Shere JA, Bartlett KJ, Kaspar CW Longitudinal study of Escherichia coli O157:H7 dissemination on four dairy farms in Wisconsin. Appl. Environ. Microbiol. 64: LeJeune JT, Hancock DD, Wasteson Y, Skjerve E, Urdahl AM Comparison of E. coli O157 and Shiga toxin-encoding genes (stx) prevalence between Ohio, USA and Norwegian dairy cattle. Int. J. Food Microbiol. 109: Cobbold R, Desmarchelier P A longitudinal study of Shiga-toxigenic Escherichia coli (STEC) prevalence in three Australian dairy herds. Vet. Microbiol. 71: Rice EW, Johnson CH Short communication: survival of Escherichia coli O157:H7 in dairy cattle drinking water. J. Dairy Sci. 83:

41 LeJeune JT, Besser TE, Hancock DD Cattle water troughs as reservoirs of Escherichia coli O157. Appl. Environ. Microbiol. 67: D Amico DJ, Donnelly CW Microbiological quality of raw milk used for small-scale artisan cheese production in Vermont: effect of farm characteristics and practices. J. Dairy Sci. 93: Robinson SE, Wright EJ, Hart CA, Bennett M, French NP Intermittent and persistent shedding of Escherichia coli O157 in cohorts of naturally infected calves. J. Appl. Microbiol. 97: Hancock DD, Besser TE, Rice DH, Herriott DE, Tarr PI A longitudinal study of Escherichia coli O157 in fourteen cattle herds. Epidemiol. Infect. 118: Garber L, Wells S, Schroeder-Tucker L, Schroeder-Tucker L, Ferris K Factors associated with fecal shedding of verotoxin-producing Escherichia coli O157 on dairy farms. J. Food Prot. 62: Schouten JM, Graat EAM, Frankena K, van de Giessen AW, van der Zwaluw WK, de Jong MCM A longitudinal study of Escherichia coli O157 in cattle of a Dutch dairy farm and in the farm environment. Vet. Microbiol. 107: Stanford K, Croy D, Bach SJ, Wallins GL, Zahiroddini H, McAllister TA Ecology of Escherichia coli O157:H7 in commercial dairies in southern Alberta. J. Dairy Sci. 88:

42 Murinda SE, Nguyen LT, Ivey SJ, Gillespie BE, Almeida RA, Draughon FA, Oliver SP Prevalence and molecular characterization of Escherichia coli O157:H7 in bulk tank milk and fecal samples from cull cows: a 12-month survey of dairy farms in east Tennessee. J. Food Prot. 65: Mechie SC, Chapman PA, Siddons CA A fifteen month study of Escherichia coli O157:H7 in a dairy herd. Epidemiol. Infect. 118: Griffin PM, Tauxe RV The Epidemiology of infections caused by Escherichia coli O157:H7 other enterohemorrhagic Escherichia coli and the associated hemolytic uremic syndrome. Epidemiol. Rev. 13: USDA-APHIS-VS An update: Escherichia coli O157:H7 in humans and cattle. USDA Animal and Plant Health Inspection Service, Veterinary Service, Centers for Epidemiology and Animal Health. Fort Collins, CO. 94. Fernández D, Rodríguez EM, Arroyo GH, Padola NL, Parma AE Seasonal variation of Shiga toxin-encoding genes (stx) and detection of E. coli O157 in dairy cattle from Argentina. J. Appl. Microbiol. 106: Hancock DD, Besser T, Lejeune J, Davis M, Rice D The control of VTEC in the animal reservoir. Int. J. Food Microbiol. 66:

43 Rhoades JR, Duffy G, Koutsoumanis K Prevalence and concentration of verocytotoxigenic Escherichia coli, Salmonella enterica and Listeria monocytogenes in the beef production chain: a review. Food Microbiol. 26: Barkocy-Gallagher GA, Arthur TM, Rivera-Betancourt M, Nou X, Shackelford SD, Wheeler TL, Koohmaraie M Seasonal Prevalence of Shiga toxin-producing Escherichia coli, including O157:H7 and non-o157 serotypes, and Salmonella in commercial beef processing plants. J Food Prot. 11: Sargeant JM, Sanderson MW, Smith RA, Griffin DD Associations between management, climate, and Escherichia coli O157 in the faeces of feedlot cattle in the Midwestern USA. Prev. Vet. Med. 66: Rasmussen MA, Casey TA Environmental and food safety aspects of Escherichia coli O157:H7 infections in cattle. Crit. Rev. Microbiol. 27: Williams KJ, Ward MP, Dhungyal OP. Daily variations in Escherichia coli O157 shedding patterns in a cohort of dairy heifers at pasture. Epidemiol. Infect. in press. DOI: Menrath A, Wieler LH, Heidemanns K, Semmler T, Fruth A, Kemper N Shiga toxin producing Escherichia coli: identification of non-o157:h7-super-shedding cows and related risk factors. Gut Pathog. 2:

44 Edrington TS, Hume ME, Looper ML, Schultz CL, Fitzgerald AC, Callaway TR, Genovese KJ, Bischoff KM, McReynolds JL, Anderson RC, Nisbet DJ Variation in the faecal shedding of Salmonella and E. coli O157:H7 in lactating dairy cattle and examination of Salmonella genotypes using pulsed-field gel electrophoresis. Lett. Appl. Microbiol. 38: Carlson BA, Nightingale KK, Mason GL, Ruby JR, Choat WT, Loneragan GH, Smith GC, Sofos JN, Belk KE Escherichia coli O157:H7 strains that persist in feedlot cattle are genetically related and demonstrate an enhanced ability to adhere to intestinal epithelial cells. Appl. Environ. Microbiol. 75: Hancock DD, Rice DH, Herriott DE, Besser TE, Ebel ED, Carpenters L V Effects of farm manure-handling practices on Escherichia coli O157 prevalence in cattle. J. Food Prot. 4: Franz E, Klerks MM, De Vos OJ, Termorshuizen AJ, van Bruggen AHC Prevalence of Shiga toxin-producing Escherichia coli stx1, stx2, eaea, and rfbe genes and survival of E. coli O157:H7 in manure from organic and low-input conventional dairy farms. Appl. Environ. Microbiol. 73:

45 870 Figure legends FIG.1. Distribution of Ct values in positive samples by sample type, for stx2 and γ-tir. MC: manure composite; filter: milk filters. Thick line: median; box: interquartile range (IQR); whiskers: quartile +/- 1.5 IQR (nominal data range); circles: data points beyond the nominal data range. Higher Ct values correspond to lower concentrations in the samples. FIG. 2. Relative abundance of whole-sample potential E. coli pathogenicity classes in dairy farm matrices. Data for Farms A, B, and C are shown in panels A, B, and C respectively. FIG. 3. Cycle threshold (Ct) categories over time in fecal samples, for stx2 and γ-tir. Downloaded from on June 19, 2018 by guest 45

46 881 Tables TABLE 1. Prevalence of virulence factors by farm and sample type Virulence factor Overall farm Fecal Manure Composite Sample type Water a Feed Milk Milk filter No. samples Farm A Farm B Farm C eae prevalence b Farm A 63.8% 66.2% 89.3% 36.4% 62.6% 6.8% 80.6% Farm B 58.3% 64.3% 91.6% 22.2% 58.9% 4.3% 42.2% Farm C 70.1% 76.3% 90.8% 26.4% 60.7% 13.4% 76.2% stx1 prevalence Farm A 29.3% 29.5% 63.0% 11.6% 24.3% 0.8% 35.2% Farm B 42.2% 47.3% 83.9% 6.5% 32.9% 0.5% 10.1% Farm C 40.5% 41.9% 76.5% 8.5% 29.8% 1.9% 30.6% stx2 prevalence Farm A 28.0% 30.1% 55.8% 7.4% 20.0% 1.1% 12.6% Farm B 43.5% 53.2% 68.7% 11.8% 25.6% 0.5% 6.7% Farm C 41.7% 47.1% 71.8% 7.5% 13.1% 0.7% 23.0% γ-tir prevalence c Farm A 13.5% 13.8% 30.8% 4.1% 2.6% 0.0% 15.2% Farm B 12.6% 11.2% 41.6% 3.2% 12.4% 0.0% 0.0% Farm C 21.2% 21.4% 43.9% 4.3% 6.9% 0.7% 19.8% a Water data in this table refer to trough water, across all age groups. b For the analysis shown in this table, samples with Ct values > 35 were considered non-detects c The number of samples analyzed for γ-tir was 5756, 4193, and 2498 for Farm A, B, and C respectively, since in the first study year γ-tir was not tested in all samples. 46

47 TABLE 2. Highest-pathogenicity E. coli strain shed by individual cows. E. coli classes were ranked from lowest pathogenicity ( non-lee non-stx E. coli NLNS) to highest (p-o157) E. coli classification NLNS LNS STEC p-ehec p-o157 Total Farm A 2.3% 6.9% 26.6% 55.0% 9.2% 1037 (21.0%) a (33.8%) (26.0%) (17.8%) (1.4%) (4522) Farm B 0.0% 1.0% 13.4% 67.7% 17.9% 403 (12.4%) (21.4%) (40.4%) (24.0%) (1.8%) (2587) Farm C 0.7% 2.8% 21.2% 64.6% 10.6% 424 (12.7%) (28.0%) (25.9%) (31.2%) (2.3%) (1548) a Values in parenthesis represent the percent of all fecal samples classified as potentially containing each E. coli category (see also Table S2). Strain categories to the right (higher potential pathogenicity) supersede categories to the left, when criteria for both are met. Not all fecal samples could be classified. 47

48 897 Figures Downloaded from on June 19, 2018 by guest FIG.1. Distribution of Ct values in positive samples by sample type, for stx2 and γ-tir. MC: manure composite; filter: milk filters. Thick line: median; box: interquartile range (IQR); whiskers: quartile +/- 1.5 IQR (nominal data range); circles: data points beyond the nominal data range. Higher Ct values correspond to lower concentrations in the samples. 48

49 FIG. 2. Relative abundance of whole-sample potential E. coli pathogenicity classes in dairy farm matrices. Data for Farms A, B, and C are shown in panels A, B, and C respectively. 49

50 FIG. 3. Cycle threshold (Ct) categories over time in fecal samples, for stx2 and γ-tir. 50