Laboratory tests for bovine respiratory bacteria and antimicrobial resistance in commercial feedlot cattle: comparing culture, long-read metagenomics, and recombinase polymerase amplification

<h4>Introduction</h4>The risk to humans and animals from antimicrobial resistance (AMR) has increased the emphasis on antimicrobial stewardship in food animal agriculture. Current stewardship recommendations include increasing diagnostic laboratory testing to inform antimicrobial use for bovine respiratory disease (BRD) management in beef feedlot production, yet the performance of newer molecular and sequencing-based diagnostic tests in commercial settings remains poorly characterized.<h4>Methods</h4>Using nasopharyngeal swabs collected from commercial feedlot calves as part of Canadian surveillance, this study evaluated diagnostic laboratory testing approaches for detecting key bacterial BRD pathogens (<i>Mannheimia haemolytica, Pasteurella multocida, Histophilus somni</i>, and <i>Mycoplasmopsis bovis</i>) and associated AMR genes. Bayesian latent class models (BLCMs) were applied to compare traditional culture and antimicrobial susceptibility testing (AST) or qPCR with long-read metagenomic sequencing and recombinase polymerase amplification (RPA). Differences in detection of target bacteria and phenotypic or genotypic AMR were assessed across the early feeding period and between age cohorts.<h4>Results</h4>This represents the first large-scale field evaluation of a recently developed, long-read metagenomic sequencing protocol implemented by a commercial laboratory for detecting BRD bacteria and AMR in respiratory samples (<i>n</i> = 760) collected by private veterinarians from western Canadian beef feedlots. Detection patterns for BRD bacteria and AMR using culture/AST and metagenomics were often similar between fall-placed calves and yearlings, but with differences from RPA. Detection of BRD bacteria had low sensitivity (< 65% for most organisms/tests), but higher specificity (>90% for all organisms/tests). Detection of macrolide and tetracycline resistance had low but variable sensitivity, with higher estimates for AST compared to metagenomics and RPA, and higher but variable specificity (>90% for most resistance outcomes/tests). Despite not using any targeted enrichment, metagenomic sequencing detected <i>M. bovis</i> although with a sensitivity lower than qPCR or RPA. Estimates of predictive value were most informative across the largest range of prevalence for AST, followed by metagenomics and then RPA.<h4>Discussion</h4>This work demonstrates the potential for large scale implementation of long-read metagenomic sequencing to support antimicrobial stewardship and AMR surveillance for feedlot cattle. The estimates of clinical diagnostic performance and predictive values provide evidence-based guidance for three different laboratory tests for BRD management.