Longitudinal analysis and characterization of Escherichia coli O157:H7 shedding in dairy cattle in northern Colorado

Abstract

Escherichia coli O157:H7 (STEC) is an enterohemorrhagic Gram-negative bacteria that is a common source of foodborne illness around the world. Annually, O157 is responsible for approximately 100,000 cases, 3,000 hospitalizations, and 90 deaths in the United States, and has been diagnostically confirmed on every continent except for Antarctica. Dairy cattle serve as asymptomatic carriers of the O157 bacteria, maintaining a continuous cycle of reinfection through their environment, and have been implicated as a potential source of contamination of the food chain. Gathering data on prevalence and shedding cycles of O157 in dairy cattle can provide insight into the scope of the problem and potential mitigation strategies. The primary objective of this study was to investigate the association between shedding status on a randomly selected day and- shedding on subsequent consecutive days (n=4), daily proportions and patterns of shedding, and how shed status on one day affects shed status on the next day. Two local Northern Colorado dairies were selected for study. Fecal samples were taken from 25 cows from Dairy A and 49 cows from Dairy B and tested for presence of the O157 pathogen. Based on those results, twenty cows from each dairy were randomly chosen for the study, with 10 “shedders” (i.e. cows that tested positive for O157 on Day 1) and 10 “non-shedders” (i.e. cows that tested negative for O157 on Day 1) selected from each dairy for a total of forty study subjects. The cows were then resampled once daily for an additional four days, testing for rfb, stx1, stx2, and eae genes as well as collecting overall health information. Health information variables were dichotomized based on scoring systems and logistic regression, generalized linear models, and generalized linear mixed models were used for analysis of research questions. Our study had three main aims and five research questions of interest. Our first aim was to analyze overall shedding events, split into two research questions. First we wanted to know if shedding status on Day 1 was associated with shedding on any subsequent day. We used a logistic regression model with any subsequent shedding as the outcome and Day 1 shedding status, dairy, parity, temperature, days in milk, body condition score, hygiene score, and fecal score as the covariates. Next, we wanted to know what risk factors were associated with cumulative days of shedding. For this question we used a generalized mixed model with a poisson regression. The count of total shedding days was used as the outcome variable and Day 1 shedding status, dairy, parity, temperature, days in milk, body condition score, hygiene score, and fecal score were the covariates. Additionally, we aimed to analyze day-to-day shedding patterns within the cattle cohort so see if shedding status on one day was associated with shedding status on the next day. First we used a generalized linear mixed model to compare paired days, specifying Day 1 vs Day 2, Day 2 vs Day 3, Day 3 vs Day 4, and Day 4 vs Day 5. The outcome variable was daily shedding status and the primary risk factor was shedding status on the stated previous day, with additional variables including Day 1 shedding status, dairy, parity, temperature, days in milk, body condition score, hygiene score, and fecal score were the covariates. We then used a generalized linear model with a logit link to assess the overall association between day-to-day shedding patterns averaged over the five-day study period, with the outcome variable as daily shedding status. The primary risk factor was shedding status on the previous day, with additional variables including Day 1 shedding status, dairy, parity, temperature, days in milk, body condition score, hygiene score, and fecal score were the covariates. Our last research question aimed to assess the associations between our risk factors of interest and daily shedding status, as well as daily shedding patterns. We used a generalized linear model with a logit link to model risk factor associations, with the outcome variable being daily shedding status and the risk factor variables including Day 1 shedding status, dairy, parity, temperature, days in milk, body condition score, hygiene score, and fecal score. We then used proportion testing to assess the differences in proportions of gene and shedding prevalence between Day 1 Shedders and Day 1 Non-Shedders. Initial shedders had a higher proportion of daily shedding than non-shedders during every sample day, 60% vs 35% on Day 2, 60% vs 45% on Day 3, 50% vs 30% on Day 4, and 45% vs 35% on Day 5, however none of these were statistically significant. Shedders similarly also had a higher overall prevalence of targeted O157 genes than Non-Shedders; 20% vs 10% for Stx1, 35% vs 30% for Stx2, and 30% vs 20% for eae. There were no significant differences in gene prevalence between cows from Dairy A and cows from Dairy B for Stx1 or eae, but there was for Stx2; 15% for both groups for Stx1, 25% vs 40% for Stx2, and 25% for both groups for eae. Cows in the Shedder cohort were twice as likely to shed O157 on any subsequent sampling day than non-shedders based on logistic regression analysis (OR 2.0, 95% CI: 1.1,3.8). Day 1 shedding status (p <0.0001), fecal score >3 vs 3 (p 0.02), and temperature (p 0.04) were significantly associated with an increase in cumulative days of shedding. Day 1 shedding status was also a significant predictor of daily shedding status (OR: 1.7, 95% CI: 1.1,2.5). Interestingly, shedding status on one day was not significantly associated with shedding status on the next day, whether looking at specific days (Day 1 vs Day 2- OR: 1.6, 95% CI: 0.4,2.5; Day 2 vs Day 3- OR: 1.5, 95% CI: 0.2,1.8; Day 3 vs Day 4- OR:1.8, 95% CI: 0.6,4.0; Day 4 vs Day 5- OR: 1.6, 95% CI: 0.3,2.2) or averaged over the 5 day study period (OR: 1.5, 95% CI: 0.9,2.3). Overall, we found inconsistent and transient shedding patterns among all of our cohorts, which is similar to findings in past literature. Day 1 shedding status was the only variable consistently found to be associated with any subsequent shedding. Although Day 1 Shedders had a higher daily proportion of shedding throughout the entire study period than Day 1 Non-Shedders, these results were not statistically significant. Past literature has said that shedding cycles likely last between two and six days, but we found that shedding status on one day was not associated with shedding status on the next day, whether looking at pair of days or averaged over the five-day period. The inconsistency in our results calls in to question whether shedding patterns are truly transient acts or whether the sampling methods used potentially misclassify Shedders as Non-Shedders.