Investigation of Diagnostic Sampling Strategies in Feedlot Calves Comparing Traditional Culture-based Analysis to Metagenomic Tools for the Identification of Bovine Respiratory Disease (BRD) Pathogens and their Antimicrobial Resistance Status

Bovine respiratory disease (BRD) remains one of the most pressing challenges in feedlot production systems and is the leading indication for injectable antimicrobial use (AMU). Accurate detection of bacterial pathogens and their antimicrobial susceptibility profiles is central to responsible AMU, yet a practical sampling strategy that enables proactive, pen-level decision-making ahead of clinical disease is lacking. Additionally, culture and antimicrobial susceptibility testing (AST) remain the traditional diagnostic standard but have inherent limitations including selective growth and incomplete characterization of resistance potential. Long-read metagenomic sequencing offers a culture-independent alternative capable of simultaneously detecting bacteria and antimicrobial resistance genes (ARGs), but its diagnostic performance and field utility for BRD management are unknown. This thesis presents an exploratory evaluation of diagnostic sampling strategies in feedlot calves by comparing traditional culture and AST with long-read metagenomic sequencing for identifying BRD pathogens (Mannheimia haemolytica, Pasteurella multocida, Histophilus somni) and their resistance profiles. The work is centered around five objectives. First, we aimed to describe pen-level prevalence and variability of BRD pathogens and AMR within and between pens during the early feeding period. Second, we sought to determine how many calves per pen must be sampled to estimate pen-level prevalence of bacteria and AMR. Third, we evaluated whether culture and AST results obtained at feedlot arrival and 13 days on feed (1 DOF and 13 DOF) could predict future BRD treatment risk and the culture and AST results obtained at first BRD treatment. Fourth, we compared the diagnostic accuracy of culture and AST to long-read metagenomic sequencing for detecting BRD bacteria of interest and AMR using Bayesian latent class models. Finally, we examined whether long-read metagenomic sequencing results obtained early in the feeding period (1 DOF and 13 DOF) were predictive of later BRD risk and phenotypic resistance at the time of BRD treatment. Collectively, the studies in this thesis provide foundational evidence that early, targeted sampling, whether based on culture and AST or long-read metagenomic sequencing, can enhance understanding of BRD pathogen and AMR dynamics, support pen-level diagnostic strategies, and inform more strategic and prudent AMU decisions in feedlot cattle ahead of clinical disease.