An official website of the United States government.

Official websites use .gov
A .gov website belongs to an official government organization in the United States.

Secure .gov websites use HTTPS
A lock ( ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

Application Of Novel Network Analyses For Predisposing Risk Factors For Antibiotic Resistant Bacteria In Cull Dairy Cattle (Usda-Grant11807521)


<p>Long-term goal: our long-term goal for this proposed project is to protect animal health and improve food safety by identifying management and host risk factors for the emergence and persistence of antibiotic resistant bacterial populations in cull dairy cattle.
<br>Specific objectives: our specific objectives are
<li>to determine the prevalence of antibiotic resistant enteric bacteria (generic E. coli and Enterococcus, gram negative and positive, ubiquitous indicator enteric bacteria) in cull dairy cattle;</li>
<li>to determine the profiles of antibioticresistance (single and multiple-drug resistant) of bacteria in cull dairy cattle; and</li>
<li>to implement innovative computational network approaches (see above) to identify clusters of risk factors (such as specific performance or clinical conditions leading to culling, antibiotic regimes and medical treatment, inadequate manure management of the cull pen, and the absence or deficient farm-level training for staff who administer the antibiotics etc.) leading to single and multiple-drug resistant enteric bacteria in cull dairy cattle.</li></ol></p>

More information

<p>Sampling strategy: Our goal is to randomly sample 400 cull dairy cattle from 20 herds in dairies from the San Joaquin Valley region. Assuming a 10% baseline prevalence for antibiotic resistant enteric bacteria, a sampling fraction of 25% for our key risk factors, and setting alpha=5%, this sample size will generate ~90% power to detect factors that elevate the odds ≥3.0 for the occurrence of antibiotic resistant bacteria. Fecal sampling will be conducted across 12 months, with 5 randomly selected cull dairy cattle sampled per season per herds. Fecal samples will be collected per rectum by trained personnel and transported to laboratory in refrigerated conditions. Upon arrival at the laboratory, samples will be stored in refrigerator (4°C) before analysis. Questionnaire: For each sampling event a standardized questionnaire will be administered at each herd. The questionnaire will include predisposing factors with a combination of 1) specific performance or clinical conditions that lead to a cull decision; 2) antibiotic regimes and medical treatments in the past 4 months; 3) management or hygiene of the sick or cull pen; 4) farm-level training for staff who administer the antibiotics; 5) feeds; and 6) environmental factors. Detection and isolation of E. coli and Enterococcus: Fecal samples will be processed within 24 h after collection. Up to 5 g of fecal sample will be dispersed and homogenized in 40 ml PBS in 50 mL centrifuge tubes. Fecal solutions will be 10-fold serially diluted (10-1 to 10-6) and 100 μL of each dilution will be streaked onto ChromAgar EC agar plate for detection of E. coli and mEI agar plate for detection of Enterococcus. E. coli plates will be incubated at 37°C for 2 h followed by incubating at 44.5°C for 22 h while Enterococcus plates will be incubated at 41.0°C for 24 h. Presumptive E. coli and Enterococcus colonies will be confirmed by biochemical tests, including Triple Sugar Iron, Urea, and Simmons Citrate agar for E. coli and Brain Heart Infusion Agar, Brain Heart Infusion Broth (BHIB), BHIB with 6.5% NaC1, and Bile Esculin Agar for Enterococcus. Confirmed colonies will be re-streaked to ChromAgar EC agar and mEI agar for isolation of bacteria respectively and 2 colonies from each positive sample will be banked at -80°C for each bacterium. Based on our experience of detection and isolation E. coli and Enterococcus from dairy samples, we anticipate that 300 isolates of E. coli and 300 isolates of Enterococcus will be obtained. Antibiotic resistance assay: Bacterial susceptibility to antibiotics will be determined using the Minimum Inhibitory Concentration (MIC) method. Briefly, 3 to 5 colonies of freshly retrieved E. coli and Enterococcus will be inoculated into 4 mL of demineralized water. Turbidities of the inoculated solution will be measured using a spectrometer and adjusted to 0.5 McFarland turbidity standards. Ten μL of the solution will be transferred into Muller-Hinton broth to yield a concentration of 105 CFU/mL. Fifty μL of the E. coli and Enterococcus solution will be inoculated into each well of the G- and G+ Sensititre® plates (Trek Diagnostic Systems Inc.) respectively. Plates will be incubated at 37°C for 24 h for E. coli and at 41°C for 24 h for Enterococcus. The Clinical and Laboratory Standards Institute (CLSI) recommended reference E. coli strains (ATCC 25922, ATCC35218) and Enterococcus strain (ATCC29212) will be used for quality control. The MIC values will be the lowest concentration of antibiotics that inhibits visible growth of bacteria. For E. coli, the G- plates screen for Amikacin, Amoxicillin/Clavulanic Acid, Ampicillin, Cefoxitin, Ceftiofur, Ceftriaxone, Chloramphenicol, Ciprofloxicin, Gentamicin, Kanamycin, Nalidixic Acid, Streptomycin, Sulfisoxazole, Tetracycline, and Trimethoprim/Sulfamethoxazole. For Enterococcus, the G+ plates screen for Chloramphenicol, Ciprofloxacin, Daptomycin, Erythromycin, Gentamicin, Kanamycin, Lincomycin, Linezolid, Nitrofurantoin, Penicillin, Quinupristin/dalfopristin, Tetracycline, Tigecycline, Tylosin tartrate, and Vancomycin.Computational Network Approach: We will use our newly developed data-driven computational techniques and specifically a combination of Percolation and Conductance (PC) (Fushing et al., 2011a; Fushing et al., 2011b), Data Cloud Geometry (DCG) (Fushing et al., 2013; Fushing and Chen, 2012; Fushing and McAssey, 2010) and Data Mechanics (DM) (Fushing and Chen, 2014REFS). The authenticity of computed pattern information and inferences are theoretically and practically developed and confirmed (Fushing and McAssey, 2010). PC can reliably compute not only a coherent tree hierarchy on study subjects, but also a tree hierarchy on antibiotics. This novel approach will allow us to first characterize the phylogenetic structure of antibiotic resistance profiles of E. coli and Enterococcus respectively based on degrees of relatedness of antibiotic resistance patterns using the PC estimating procedures. Next, we will determine which combination or cluster of risk factors in dairy herds correlate with specific clusters of highly related profiles of antibiotic resistance of E. coli and Enterococcus (Fig. 1) using DCG and the shuffling algorithms of DM, similar to our related work underway (Fushing et al., 2013; Vanderwaal et al., 2013; Vanderwaal et al., 2014; Fushing and Chen, 2014). Coupling these two trees together using these techniques makes a powerful platform for identifying critical information, including which subject clusters are susceptible or resistant to what antibiotics, and the risk factors underlying the subject-vs.-antibiotic interacting patterns (Fig. 1). For example, assume the community network for multiple-drug resistant bacteria from cull dairy cattle resembles the data phylogenetic tree shown on the X-axis and phylogenetic tree of risk factors shown on the Y-axis in Figure 1, which in this theoretical example exhibits several highly related clusters from isolates in X-axis and related clusters of risk factors in Y-axis. The cluster referred to by the blue arrow in the middle of the figure represent a group of cattle that exhibit similar patterns of antibiotic resistance and exposed to similar clusters of risk factors. Thus risk factors (such as antibiotic regimes, medical treatment, cull decision, or pen management etc.) leading to single or multiple-drug resistant (such as resistant to Tetracycline, Kanamycin, and Penicillin) will be identified.The proposed methods described above have significant advantages over more conventional methods. For example, hierarchical clustering procedures are often used to construct phylogenetic trees of pathogen genetic relatedness, but such clustering methods can result in biased and thus invalid calculations of the network topology or geometric structure of the relatedness tree (Fushing et al. 2013; Fushing and McAssey, 2010). Although conventional parametric methods can be used such as log-linear regression or polychotomous logistic regression to link individual risk factors to specific patterns of antibiotic resistance, we have found limitations to this approach as it does not allow the identification of related sets or clusters of factors and therefore can lead to limitations in identifying efficacious prevention strategies based upon interdependent and collinear risk factors (Berge et al., 2005; Jan et al., 2012; Kirk et al., 2005; Stoddard et al., 2009). Adoption of our new computational network techniques will allow us to better identify clusters of risk factors that underlie the observed antibiotic resistant patterns in cull dairy cattle.</p>

University of California - Davis
Start date
End date
Project number
Accession number