Abstract
Escherichia coli O157 is a foodborne pathogen that can be transmitted by contaminated ground beef and is shed naturally in cattle feces. Recent reports indicated that feeding distillers' grains (DG) to cattle increased fecal shedding and prevalence of E. coli O157. In Minnesota, feeding DG with solubles (DGS) to livestock became widespread within the last 10 years, but there is no report about the prevalence of E. coli O157 in beef cattle in this state. This study was undertaken to survey the fecal prevalence of E. coli O157 in cattle fed diets containing DG and its association with environmental conditions and management practices. Fecal samples were collected from three feedlots during a 1-year period. All animals in those feedlots were fed different DGS levels. E. coli O157 presence was determined using a combination of enrichment, immunomagnetic separation, plating onto sorbitol MacConkey agar, and confirmation of isolates by immunoassay and multiplex virulence genes polymerase chain reaction analysis. Overall, E. coli O157 was confirmed in 9.7% of samples. Prevalence during summer was 30% and declined to less than 10% the rest of the year. In animals grouped by dietary DGS concentration, no significant difference in prevalence (12.0 and 5.5%) was detected between the low and the high average groups (less and more than 20%). Previous feeding of DGS before arriving to the feedlot also had no influence on fecal prevalence. The presence of several interacting variables, uncontrolled in a real-life feedlot environment, was the likely reason for our observation and suggested that at the levels studied, DGS had no effect on the STEC O157 prevalence in cattle populations.
Introduction
I
The factors influencing the presence of this pathogen within a herd are still unclear, although some factors, such as age of the animals (Mechie et al., 1997; Van Donkersgoed et al., 1999), seasonality (Chapman et al., 1997; Garber et al., 1999), diet, and geographic location have been identified. Also, different studies monitoring E. coli O157:H7 in feces reported large variations in the number of positive cattle, but observed a small fraction of animals that had more than 104 colony-forming units/g of feces and they were referred to as super-shedders (Omisakin et al., 2003; Fegan et al., 2005).
The cattle feed composition has been actively investigated as a risk factor for STEC O157 (Callaway et al., 2009). Distillers' grains with solubles (DGS), an ethanol fermentation byproduct, have been identified as a dietary risk factor. Because of the increase in ethanol demand, DGS has become abundant and used extensively as a component of cattle feed. A link between increased STEC O157 prevalence and DGS was first suggested by Synge et al. (Synge et al., 2003). Several studies have been performed on animals under controlled feeding and environmental conditions, either naturally shedding or artificially inoculated (Jacob et al., 2008b; Wells et al., 2009; Jacob et al., 2010), and on STEC O157 naturally present or inoculated in fecal slurries (Varel et al., 2008; Varel et al., 2010; Yang et al., 2010). These studies suggested correlations between feeding DGS and increased prevalence of the pathogen. However, a recent study comparing DGS to dried rolled corn failed to find any connection with STEC O157 prevalence (Jacob et al., 2009).
As a major ethanol producer, Minnesota has seen a large increase in the use of DGS. Interestingly, several major outbreaks of STEC O157 have originated in this state, but very little is known about the prevalence of this pathogen in local cattle populations and its possible link with DGS use. This work was aimed at estimating the fecal prevalence of E. coli O157 in beef cattle in Minnesota and addressing the purported relation with farm management practices and dietary DGS.
Materials and Methods
Sample collection
Three Minnesota beef cattle feedlots, with a combined one-time capacity of approximately 7000 animals, were surveyed from October 2009 to October 2010 at least once a month. The sampling period was subdivided in four seasons: fall, from September to December 2009, and from September to October 2010; spring, from March to June; summer, from June to September; and winter, from December to March.
The feedlots used open, enclosed, and slatted-floor facilities and housed cattle bought from different sources to be finished for slaughter. Feedlots were commercial animal feeding operations, and the herds sampled comprised animals from calves with an entry weight of at least 300 kg to grown yearlings weighing 600–700 kg. Animals included ranged from those that had been fed their diet for 5 days to as long as 220 days. Depending on the pen size (number of animals in a pen), 4–12 pens were sampled at each site, pens had a mean of 138 (standard deviation [SD]±76) animals, and an average of 5% (SD±2%) of the cattle present in each pen were sampled. To ensure sampling from single individuals, fecal samples were randomly collected from the feedlot floor immediately after defecation.
Demographic information (i.e., animal gender, age, source prior to entering the feedlot) and herd management practices (i.e., facility type, ration, and production system) from each cattle load were collected using a questionnaire. The samples collected were 1376 and the total number of heads were categorized by gender, age, DGS concentration (dry matter basis, <20% and ≥20%), type of facility, type of DGS, and sources prior to entering the feedlot (Tables 1 and 2). Fecal samples were transported to the laboratory, stored at 4°C and tested within 24 h after collection.
Samples collected (total number of animals at the time of sample collection).
The total cattle population in the three feedlots at the time of sampling is shown in parentheses.
Isolation of E. coli O157 by immunomagnetic separation
Sample portions (10 g) were mixed with 90 mL of modified EC broth (20) supplemented with novobiocin (20 μg/mL; Sigma Chemical Co., St. Louis, MO) and blended by stomaching. Samples were incubated for 2 h at 37°C; duplicates of 1 mL of cultures were subjected to immunomagnetic separation according to Schamberger et al. (2004). The immunomagnetic beads were suspended in 200 μL of the washing solution, spread plated onto sorbitol MacConkey agar (Neogen, Inc., Lansing, MI; supplemented with cefixime [50 μg/L; Lederle Laboratories, Pearl River, NY] and potassium tellurite [25 mg/L; Sigma]), and incubated at 37°C for 18–24 h. Typical clear colonies were tested with an O157 immunoassay (RIM E. coli O157 Latex Test, Remel, Lenexa, KS).
Identification of virulence factor genes by multiplex PCR
Positive agglutination assay isolates were tested by multiplex PCR for the presence of virulence genes. The primers were specific for Shiga-like toxins (stx1 and stx2), intimin (eaeA), and hemolysin (hlyA) to confirm an O157 isolate. The PCR primers and conditions are reported in Table 3.
The thermal cycling consisted of 35 cycles, 1 min at 95°C; 2 min annealing at 65°C for the first 10 cycles, decrementing to 60°C by cycle 15; and 1.5 min of elongation at 72°C, incrementing to 2.5 min from cycles 25 to 35. Polymerase chain reaction reaction mixtures were electrophoresed on 2% agarose gels and stained with ethidium bromide. This protocol was based on that of Paton and Paton, 1998a.
Statistical analyses
Logistic regression was used to model the log odds of a sample testing positive for E. coli O157 as a function of predictor variables: percentage DG in feed (16.6, 23.8, missing value), operation type (open versus confined), prior history of distillers grains in diet (yes, no, mixed), age (calf or yearling), days on feed (over/under 60 days, missing value), sex (heifer or steer), type of operation (natural, conventional), and season (fall, winter, spring, summer). Random effects were incorporated into the model to adjust for variation between feedlots and pens within feedlots. The variance component for feedlot variation did not differ from zero. However, there was evidence of pen-to-pen variation within feedlot, and this random-effect component was included in our final model. Logistic regression modeling with random effects used the Glimmix Procedure of SAS (SAS Institute Inc., Cary, NC) with the binomial link, and the Kenward-Roger adjustment to the covariance matrix of parameter estimates. Interacting predictor effects were assessed through models suggested by forward-and-backward model selection approaches. Model results are presented in Table 4 with odds ratios for each effect relative to a specified comparison category. Confidence intervals for each estimated odds ratio (OR) are also provided to assess the uncertainty of the estimated OR. Estimated ORs clearly above or below 1.0 provide evidence that the factor level of interest differs statistically from the comparison category. Residuals, leverages, and influence diagnostics were examined during model selection to guide the form of the final model.
Odds ratio defines the probability of finding a positive sample relative to the category with a probability of 1. These ratios are model-adjusted.
Significant variable when confidence interval (95%) does not encompass 1.00.
Concentration of distillers' grains with solubles (DGS) in the diet. Proc Univariate was used to distribute.
Open is defined as having access to an outside lot.
Yearling defined as a bovine older than 1 year of age.
Results
Sampled animals distribution
Over a year span, the animals sampled with repetition were 1376, representing a total population of 27,906 head covering a total of 213 pens. The animals were distributed among three feedlots in West Minnesota (Feedlots 1, 2, and 3) and sampled in similar percentages: 33.9%, 38.9%, and 27.2% of the samples, respectively (Table 1). The sex distribution was 86.3% steers and bulls and 13.7% heifers. The majority of the samples, 66.7%, came from year-old animals (yearlings) and the remainder (33.3%) were calves. Most of them, 72.7%, were raised in open facilities and 27.3% were raised in confined barns.
All animals in the present study were fed wet DG during the finishing. The values of DGS were estimated on the dry weight. Prior to finishing, 59.2% were pasture fed and the remainder 40.8% of the animals had previous access to DGS (Table 2). DGS levels were used to define two classes of DGS feeding levels, with cut-offs providing sufficient sample sizes per class. Two thirds of cattle were raised as conventional type, that included the use of growth hormones and antibiotics, and the rest were raised as “natural,” defined by restricted or no use of antibiotics and hormones. Over half of the sampling was performed during the fall seasons (2009 and 2010), 27.4% during spring (2010), 9.3% in summer (2010), and 8.8% in winter (2009–2010).
Prevalence of E. coli O157 in cattle and associations with management factors
The animals shedding E. coli O157 were 134 of a total of 1376 samples, resulting in an overall prevalence for STEC O157 of 9.7%. Feedlot 1 had <5% prevalence, whereas Feedlots 2 and 3 had prevalence rates of 12.3 and 12.0%, respectively (Table 5). The prevalence distribution between genders was almost twice for females (16.4%) as for males (8.7%) (Table 5), but the OR adjusted for other factors was not significant (Table 4). In the age class, yearlings had a prevalence of 11.1%, and calves of 7.0%. Yearlings were twice more likely to carry E. coli O157 than calves (Table 5), with a model adjusted OR of 2.73 (Table 4). Animals with access to DGS prior to the finishing period tested positive for STEC O157 (14.7%) more frequently than animals that did not (6.2%), but the OR value was not significant. The concentration of DGS in the diet was subdivided into two categories based on the DGS percentage expressed on dry matter (e.g., <20.0, ≥20.0%, and missing value). The lowest class had the highest prevalence, but the model-adjusted OR values were not significant. However, the highest level of DGS resulted in higher prevalence in summer (OR of 0.04 and 0.09 in the fall and spring, respectively) relative to that in other seasons (Table 4).
The natural and conventional categories had no difference in prevalence (Tables 4 and 6). Summer was the season with the highest prevalence (more than 30%), and winter had roughly half of that prevalence, which was not significant. However, the prevalence during spring and fall (10.6% and 4.9%, respectively) was significantly less than during summer (Table 4). Animals that were kept in open spaces had over twice the E. coli O157 prevalence of those that were housed in confined barns (Table 5), but this difference was not significant (Table 4). The analysis of the interaction between barn type and seasons resulted in a significant positive interaction between the use of open barns, spring, and likelihood of finding positive samples (Fig. 1).
Least-square (LS) means for the interaction between confinement and season with 95% confidence intervals. , Open feedlot; 
, Confined feedlot.
Distribution of the genotype frequencies based on Shiga toxin genes
All of the 134 STEC O157 isolates carried stx2, and more than half of the isolates (55.2%) had the combination of stx1 and stx2 and 44.8% carried stx2 only (Table 7). Genotypic frequencies by origin, gender, age, and barn type were significantly different from the general distribution observed, with p values lower than 0.01. Feedlot 1 significantly departed from the expected frequencies, with 21.7% of the isolates carrying stx1 and 2 and 78.3% only stx2. In Feedlot 2, 63.6% of the isolates had stx1 and 2, and 36.1% had only stx2. In Feedlot 3, 60.0% of the isolates had both stx1 and 2 and 40.0% had only stx2 and the difference from the general frequency was not significant (Table 7). The frequencies among male animals followed the general population pattern; however, samples from heifers and cows had a significantly more frequent occurrence of the combined stx1 and 2 genotype (76.9%). With the exception of fall, seasonal frequencies of the genotypes were all significantly different from the general distribution, with the stx2-only genotype reaching a frequency of 88.2% in winter and the combination stx1 and 2 85.0% in spring.
p<0.01. NS, not significant.
Discussion
This work is the first large cross-sectional study focusing on the prevalence of E. coli O157 in beef cattle raised in Minnesota. Even though several STEC O157 outbreaks have occurred in Minnesota and ethanol-derived DGS is used extensively in cattle feed there, only one previous study included a survey of STEC O157 prevalence in the state. In a longitudinal study, Hancock et al. grouped the Minnesota data with six other states in a “Northern” United States region with an overall prevalence for E. coli O157 of 2.5% and a feedlot prevalence of 80% (Hancock et al., 1997). However, those values were determined before the use of immunomagnetic separation. Introduction of immunomagnetic separation markedly enhanced the recovery rate for STEC O157; the majority of recent studies reported levels of STEC O157–positive cattle between 5% and 30% (Laegreid et al., 1999; Barkocy-Gallagher et al., 2003; Sargeant et al., 2003; Alam and Zurek, 2004). Seasonality has been shown to have a large impact on the prevalence of STEC O157. Not surprisingly, significant differences in O157 presence were observed during summer (31.3%), albeit statistically tied with winter. Higher prevalence of O157 in fecal samples during summer was observed in other studies (Van Donkersgoed et al., 1999; Meyer-Broseta et al., 2001; Barkocy-Gallagher et al., 2003; Smith et al., 2005).
Previous studies performed under controlled conditions and examining the effect of feeding DGS on fecal prevalence of E. coli O157 determined through culture have yielded diverse results. Dewell et al. (2005) were one of the first research groups that found feeding DGS as a risk factor for E. coli O157:H7 carriage in a beef cattle longitudinal study. Two studies using naturally infected feedlot cattle (Jacob et al., 2008c) and experimentally infected Holstein calves (Jacob et al., 2008a) reported higher prevalence of E. coli O157 when DGS was included in the diet. However, two subsequent reports did not find the same relationship (Jacob et al., 2008b; Jacob et al., 2009). Similarly, Edrington et al. did not see any influence of DGS on fecal prevalence (Edrington et al., 2010). A recent study from Wells et al. showed that levels of DGS higher than 40% can be associated with a higher prevalence, particularly when fed in seasons characterized by a lower prevalence (Wells et al., 2009). In our observations, animals fed less than 20% DGS had the highest prevalence, but, the model-adjusted OR values were not significant. However, the risk associated with DGS changed significantly across seasons, and our analysis revealed a significant interaction between the DGS percentage and the season (Table 4), in accordance with previous reports (Wells et al., 2009; Jacob et al., 2010). However, differently from that work, our results indicate that levels of DGS higher than 20% over DM of feed lead to a notably higher prevalence of STEC O157 in summer relative to any other seasons.
We also found significant positive interactions between open barns, spring months, and likelihood of finding positive samples (Fig. 1). Previous reports have found a relationship between O157 prevalence and wet and muddy pen floor conditions (Smith et al., 2001). In our case, open barns are generally drier and cleaner than confined barns. Unfortunately, we could not evaluate this interaction in a model that also contained a season by DGS concentration interaction because there are insufficient positive cases at the combinations of DGS, confinement, and season to estimate all interaction effects.
In our study, age was also a factor in the prevalence of STEC O157, with older animals (yearlings) having a higher prevalence and being almost three times more likely to carry this serotype than calves (11.1 and 7.0%, respectively, with an OR of 2.73). Previous studies have also found lower prevalence in calves than in adult animals (Cizek et al., 1999; Conedera et al., 2001), but at least two other studies found higher prevalence in calves than in adults (Gannon et al., 2002; Nielsen et al., 2002). Our work differs from all these analyses in that it compares calves to yearlings and not to full-grown adults.
Previous studies have found that stx2, the gene most frequently associated with hemolytic uremic syndrome, was the most prevalent (Padola et al., 2004; Fremaux et al., 2006; Rivas et al., 2006; Fernández et al., 2010; Rasooly and Do, 2010). In our study, we found that stx2 was present in all isolates, and 55.2% of the times it was associated with stx1 (Table 7). In accordance with the literature reporting that stx1 is rarely found alone, particularly in O157 (Galland et al., 2001; Fernández et al., 2009), no isolate carried this gene only. We also compared the seasonal variation in Shiga-toxin distribution in the isolates and found that there is a strong seasonality with an increase of the stx1/stx2 genotype in spring.
Footnotes
Acknowledgments
This work was supported by the Rapid Agricultural Response Fund from the Minnesota Agricultural Experiment Station.
Disclosure Statement
No competing financial interests exist.