Agent-based and contact network modeling applications for Escherichia coli transmission in commercial feedlot settings

Abstract

Shiga toxin-producing Escherichia coli (STEC) are recognized as a major food-borne pathogens with outbreaks, human infections, and occasional deaths associated with the consumption of contaminated foods. Cattle are recognized as a primary reservoir for STEC, though the transmission dynamics of STEC in feedlot cattle are not fully understood. Our current understanding of the transmission dynamics is dependent on longitudinal studies of naturally occurring STEC infections in cattle. A certified nonpathogenic strain of E. coli was used for a series of transmission experiments of a known source and concentration to better understand the transmission dynamics of STEC in commercial feedlot settings. The contact networks of feedlot cattle both within and between pens were characterized and an agent-based model was created to examine the influence of direct and indirect transmission routes on the pen-level prevalence of E. coli colonization. Nonpathogenic E. coli strains were used as a surrogate to assess the transmission dynamics for STEC in cattle. Initially, three verified nonpathogenic E. coli strains (O19:H+, O101:H10, and O28:H43) were orally inoculated into weaned Holstein calves. All inoculated strains were able to colonize in the gastrointestinal tract of the calves, shed in their feces, and spread to their pen environment. Pen-level area under the curve (AUC) for fecal shedding concentrations of each nonpathogenic strain were evaluated and the strain with the highest pen average fecal AUC was selected for use in two independent inoculation trials that each occurred in one independent pen of 70 feedlot cattle. E. coli strain (O28:H43) was selected for use in these two inoculation trials. Oral inoculation of five randomly selected steers from each pen occurred for five days at the start of the study period. The inoculated strain was able to colonize, shed in the feces, spread to the pen environment, and transmit between feedlot steers. These data provide baseline data on the shedding and transmission of a nonpathogenic E. coli strain in diary calves, feedlot steers, and its detection in the pen environment. These data can be furthered used as a surrogate to better understand enteric pathogen transmission, such as STEC, in commercial U.S. feedlot systems to explore effective interventions options in a real-world setting. To better inform the construction of network-based disease transmission models for cattle housed within confined-spaced systems, contact network modeling was used to quantify the contacts defined with a spatial threshold (SpTh) of 0.71 m and a minimum contact duration (MCD) of either 10, 30, or 60 seconds within three pens of feedlot cattle across consecutive years. Static, undirected, weighted contact networks were created for the full study duration and at vary timescales (24-h, 6-h, and h) to assess network heterogeneity. The influence on contact networks in feedlot cattle due to the variation in Real-Time Location System (RTLS) average tag read rate observed between the three years though the same system was used was examined. When the networks were down-scaled from higher average tag read rates to match the lower average tag read rates, the overall networks maintained similar network density and clustering, though the average edge weight between pairs of steers decreased. The high-resolution spatial and temporal contact data provided estimates for contact networks within U.S. commercial feedlot pens that can be used to better inform pathogen transmission models. Contact network analysis was used to quantify contacts and compare the resulting static, undirected, weighted contact networks created from two neighboring pens of feedlot cattle and from across the shared fenceline at varying timescales (24-h, 6-h, and h). Contacts within-pen were defined at 0.71 m SpTh with a MCD of 10, 30, or 60 seconds. “Fenceline” contacts were defined with the same SpTh and MCD durations as within-pen contacts but contained a steer from each pen and occurred within the defined fenceline area, any location reading occurring within 1 m of either side of the shared fenceline. On a full study duration, the contact networks created within each feedlot pen were densely connected. On shorter timescales (24-h, 6-h, h) the within-pen contact networks showed greater network heterogeneity in density and clustering metrics. The contact network created across the shared fenceline yielded a total network density of 17%. These findings can be used to better inform the construction of network-based disease transmission models for cattle within confined-spaced systems and additionally accounting for the transmission that occurs over man-made barriers (e.g., fencelines). Finally, an agent-based model was created in Netlogo 6.2 using empirical data to explore the influence of direct steer-to-steer contact and indirect steer-to-pen environmental contact on the pen-level prevalence of E. coli colonization. Agents in the model were defined as individual steers and three pen environment areas (front, middle, rear). Direct contact between pairs of steers in the model was defined as cumulative 10 s temporal sampling windows (TSW) aggregated on a daily (24-h) level. Indirect contact between individual steers and the three pen environment areas were defined by using a proportion of location readings for each steer in each defined pen area for a given study day. Proportions were then averaged for each individual steer by study week in the model. Colony forming units (CFU) of E. coli transmitted by direct contact were modeled as one, five, and ten percent of the average log₁₀CFU/400cm² of collected hide samples, and CFUs of E. coli per pen area were determined as one, five, and ten percent of the average log₁₀CFU/g of those pen soil surface samples that fell within that pen area for that study week. The colonization period of individual cattle within the model was defined as a weighted average of the time in study days in which all individual steers in the empirical data were positive in the digitally collected fecal samples. Preliminary results are suggestive that direct contact between animals is more important for pen-level prevalence for E. coli colonization than indirect contact with pen environment. Additional work is needed to determine the R-nought of the E. coli modeled to compare it to published values of R-nought of STEC.