Antimicrobial susceptibility of Pasteurella multocida and Mannheimia haemolytica in dairy farm

INTRODUCTION

Bovine respiratory disease (BRD) remains a costly challenge to the dairy industry, as the bacterial complex involved in BRD has shown decreased susceptibility to antimicrobials, exacerbating its effects on animal health, productivity, and recovery. In human medicine, cumulative antimicrobial susceptibility testing (CAST) is used to generate antibiograms that support empirical decision-making for antimicrobial treatment. The goal of this study was to evaluate a CAST framework for BRD in dairy cattle and factors of relevance when analyzing and reporting antibiogram data for on-farm treatment decisions. For this purpose, a repeated cross-sectional study was conducted using criteria adapted from the Clinical Laboratory Standards Institute for the collection, analysis, and presentation of CAST data. From 3 dairy farms in California, calves, heifers, and cows were sampled using deep nasopharyngeal swabs for culture and antimicrobial susceptibility testing for Pasteurella multocida and Mannheimia haemolytica. Antimicrobial susceptibility testing was performed using the minimum inhibitory concentration approach for 12 antimicrobial drugs of relevance to dairy cattle. The variability in the odds of nonsusceptible isolates was evaluated for effects of age group, farm, and season using a logistic regression model in SAS. The antimicrobial susceptibility of co-occurring P. multocida and M. haemolytica isolates were evaluated using Pearson chi-squared and Cohen’s kappa in SAS. Overall, P. multocida isolates were mostly nonsusceptible to tetracycline (74%) and spectinomycin (46%), whereas M. haemolytica isolates were frequently nonsusceptible to penicillin (43%) and tetracycline (34%). There were increased odds of nonsusceptible isolates across farms and age groups, with calves and heifers showing a higher number of nonsusceptible isolates than cows. For P. multocida, isolates nonsusceptible to tetracycline and spectinomycin were affected by age group, with calves showing greater odds of nonsusceptibility to antimicrobials. The farm effect was significantly associated with P. multocida nonsusceptible to all antimicrobials but penicillin, suggesting individual farm management practices may heavily affect profiles of nonsusceptible BRD bacteria. Similarly, nonsusceptible M. haemolytica was consistently associated with farm for macrolides and tetracyclines and with age, with heifers showing the greatest odds of being nonsusceptible when compared with cows. Season did not affect the nonsusceptibility of either pathogen, suggesting that either California seasonal patterns or other factors, such as farm practices and animal management, play a more prominent role in the change of susceptibility profile. Overall, these findings highlight the potential effects of farm and age when interpreting data from CAST reports for BRD, where these and other factors should be considered when analyzing and presenting antibiogram reports.

Antimicrobial resistance (AMR) brings significant challenges to both human and veterinary medicine, posing a risk to the success of bacterial infection treatments and a risk of poor prognosis. The increase in AMR is particularly critical in the management of bovine respiratory disease (BRD), a major cause of morbidity and mortality in the cattle industry and a major source of antimicrobial use on dairy farms. This is a multifactorial disease often associated with Mannheimia haemolytica and Pasteurella multocida, 2 primary bacterial pathogens that are commensals of the upper respiratory tract in healthy cattle. The emergence of isolates nonsusceptible to antimicrobials for these pathogens challenges the effectiveness of on-farm treatment protocols, especially with the increasing identification of multidrug-resistant isolates for relevant BRD bacteria in cattle.

Antibiograms, specifically cumulative antimicrobial susceptibility test (CAST) reports, are critical tools for guiding initial antimicrobial therapy by providing data on bacterial susceptibility patterns over time, thereby reducing empiricism in disease treatment. Although CAST is widely used in human medicine, its application in veterinary medicine, particularly in livestock, faces several challenges. These challenges stem from diverse management practices across commercial farms, including differences in animal handling, facility infrastructure, and disease treatment protocols. Moreover, the difficulty of obtaining samples from farms, the limited availability of livestock-approved antimicrobials, the costs associated with bacterial culture and susceptibility testing, and the herd-health approach in livestock management contrast sharply with the individualized care in human medicine. Thus, to successfully implement CAST data in antibiogram reports for cattle, these limitations must be addressed to enhance the validity of the and maximize its effects, ensuring a better return on investment. Nevertheless, there is currently limited information on the factors that should be considered when generating antibiogram reports for BRD in dairy cattle, underscoring the need for further evaluation.

A critical step for increasing the effects of CAST reports in a population is understanding what factors may significantly be associated with a variation in susceptibility to specific disease-causing bacteria and antimicrobial drugs. As an example, a study on beef cattle at entry into Canadian feedlots found that the production source of the animals (beef vs. dairy) was associated with differences in the odds of isolating nonsusceptible antimicrobial-resistant strains when cattle were diagnosed with BRD. Similarly, CAST reports in human medicine have identified specific factors that contribute to variability in the nonsusceptibility to antimicrobial drugs. For example, the source of the isolates has been recognized as a major factor affecting nonsusceptibility in bacterial pathogens. In contrast, limited data exists on factors significantly influencing the nonsusceptibility of BRD isolates in dairy cattle, which affects CAST result interpretation and use of antibiogram reports. Additionally, in almost all cases of respiratory disorders in dairy cattle, antimicrobial therapy is administered. However, most epidemiological studies on BRD isolates nonsusceptible to antimicrobials have focused on feedlot cattle, a distinct population with management strategies that may affect nonsusceptibility profiles.

Therefore, the main goal of this project was to evaluate an antibiogram framework for the longitudinal collection of samples to evaluate the nonsusceptibility profile of BRD bacterial pathogens and identify factors of relevance to consider for a CAST program. With that in mind, the specific aims of this study were as follows: (1) to evaluate a framework for antimicrobial susceptibility testing on M. haemolytica and P. multocida isolates from BRD cases in dairy cattle diagnosed with the disease on California dairies; and (2) to analyze the variability in the nonsusceptibility profiles of these BRD pathogens when stratified by factors such as age group, farm, and season. We hypothesized that factors such as age, farm, and season significantly affect the nonsusceptibility of BRD pathogens to antimicrobials on dairy farms and that such factors should be considered when presenting antibiogram reports.

METHODS

This study was conducted from November 2021 to October 2023, and all procedures were approved by the Institutional Animal Care and Use Committee at the University of California, Davis, under protocol #223017.

Study Design

A repeated cross-sectional study was conducted on a convenience sampling basis at 3 large (approximately 1,000 lactating cows in each) commercial dairy farms in Northern California. The sample size calculation was performed using the LRCHI method in PROC POWER from SAS 9.4 (SAS Institute Inc.) based on a 95% confidence level and 80% power. We based these analyses on the Clinical Laboratory Standards Institute (CLSI) guidelines for CAST, which recommend that only organisms with 30 or more isolates tested during an analysis period (1 yr for this study) should be included to obtain a reasonable statistical estimate of the cumulative percent susceptible rate. Thus, based on previous reports from CLSI, we expected a 40% nonsusceptibility rate to antimicrobials in isolates at an expected recovery rate of 36% to 60% for P. multocida and 20% to 21% for M. haemolytica for BRD score-positive animals compared with score-negative animals. The criteria for BRD case definition are outlined in the following sections of this manuscript. A minimum of 150 samples per year originating from dairy cattle with clinical signs of BRD, obtaining an estimated minimum of 36 and 65 isolates per year for M. haemolytica and P. multocida, respectively, were expected.

Case Definitions

Once or twice a month, a veterinarian visited each of the participating dairy herds and observed all lactating cows from the hospital pen, calves (preweaning calves), and heifers (weaned calves). All animals suspected of having BRD were marked for further examination following a standardized procedure for each age group, which consisted of clinical examination and BRD scoring, thoracic auscultation, lung ultrasound (only in calves), and collection of nasal swabs for culture and antimicrobial susceptibility testing. For thoracic ultrasound in preweaning calves, an Easi-Scan:Go ultrasound with a linear probe and a wireless BUG OLED binocular headset viewer was used (IMV Imaging Ltd.).

Calves were suspected of having BRD if they presented with BRD signs such as eye discharge, nasal discharge, cough, or abnormal respiration. Then, the evaluation was confirmed as a BRD case when the California BRD cumulative score was ≥5, with abnormal findings on thoracic auscultation or ultrasound or both. Heifers were suspected of BRD if they presented BRD signs such as nasal discharge, sunken eyes, cough, or abnormal respiration. Cases of BRD in these heifers were confirmed when their cumulative scores were ≥2 (diurnal temperature data were not used), based on the validated scoring system for postweaning dairy calves, with abnormal findings on thoracic auscultation. Cows were suspected of having BRD if they presented primary diagnostic criteria of BRD, such as cough, nasal discharge, increased respiratory effort or rate, and abnormal findings on thoracic auscultation, or secondary diagnostic criteria, such as depressed demeanor and low body condition score. A cow with at least 2 primary diagnostic criteria or 1 primary and 1 secondary was confirmed as having BRD. Only females were included in the study. Animals were excluded if they were severely depressed, apathetic, or unable to stand, as well as if there was no permanent ID or ear tag.

Sampling methods followed recommendations by the CLSI guidelines for analysis and presentation of CAST data for veterinary antibiograms, using only diagnostic isolates, and eliminating any duplicate samples, with each individual animal being only sampled once.

Sampling Procedure

Samples were collected using a single-guarded deep nasopharyngeal swab approach, as previously described by it. Briefly, samples were collected using an individual guarded sterile cotton swab (KI-3000, Kalayjian Industries, Inc., Signal Hill, CA) by restraining animals in a standing position. The animals’ nostrils were wiped clean with a single-use paper towel and subsequently disinfected with a 70% alcohol wipe before the sterile swab was inserted medioventrally into the nasal cavity until nasopharyngeal tissue was reached and rotated several times against the mucosa. Swabs were immediately placed in Amies with charcoal transport media for bacterial isolation (ACM, BD BBL CultureSwab Plus Transport Systems, Franklin Lakes, NJ) and transported in a cooler with ice to the laboratory for processing.

Bacterial Isolation and Antimicrobial Susceptibility Testing

Aerobic culture of pathogens was performed as described by it. Briefly, within 4 h after collection, samples were submitted to the California Animal Health and Food Safety Laboratory in Davis, California, for bacterial culture, speciation, and antimicrobial susceptibility testing. Each sample was cultured on 5% sheep blood agar (SBA), chocolate agar (CHOC), and MacConkey agar. The CHOC was used alongside SBA primarily to enhance the initial growth of potentially fastidious pathogens. However, scoring for colony morphology and isolation of pure colonies for subsequent analyses were consistently performed using only blood agar plate. Chocolate agar was not directly used in subsequent analyses or reporting of results. Plates were incubated for 24 h at 35 ± 2°C in 5% CO₂ with the goal of recovering M. haemolytica and P. multocida. All colonies of interest were confirmed by biochemical testing and MALDI-TOF MS (Bruker MALDI Biotyper Sirius One GP System, Bruker Daltonics, Fremont, CA) using the BDAL database version 10. Biochemical tests performed included catalase, oxidase, and Gram stain assays for initial characterization. Species-level confirmation was based on MALDI-TOF scores >2.0, per manufacturer guidelines. Culture results on the SBA plates were used to score growth into 1 of 5 options: (1) pure growth, with only 1 type of bacterial colony; (2) 1 dominant colony morphology, with other colony types present; (3) 2 dominant colony morphologies; (4) polymicrobial growth, with more than 2 different dominant colony morphologies; and no growth (NG).

Isolates were evaluated for antimicrobial susceptibility using a microbroth dilution method following the CLSI guidelines. Isolates were inoculated into Sensititer VetBovine BOPO7F plates (Sensititer Susceptibility Plates, Thermo Fisher Scientific, Waltham, MA), as recommended by the manufacturer’s instructions. The inoculum was prepared with Sensititer Sterile Water at 0.5 McFarland Standard using a nephelometer (Thermo Fisher Scientific, Waltham, MA) and mixed with cation-adjusted Mueller-Hinton broth with laked horse blood (Thermo Fisher Scientific, Sensititre, Waltham, MA) according to CLSI guidelines. Plates were incubated at 36°C for 24 h and subsequently read using a Sensititer OptiRead Automated Fluorometric Plate Reading System. Quality control was routinely performed weekly according to CLSI guidelines. The quality control organisms included Escherichia coli ATCC 25922, Staphylococcus aureus ATCC 29213, and Mannheimia haemolytica ATCC 33396, with results required to fall within CLSI-specified ranges to confirm assay validity. The BOPO7F plate was used to test the minimum inhibitory concentrations (MIC) on the following 12 antimicrobial drugs that have MIC breakpoints (clustered by antimicrobial classes) based on current CLSI breakpoints aminocyclitols (spectinomycin [tested range = 8–64 µg/mL; MIC breakpoint for both isolates to be considered nonsusceptible is ≥64 µg/mL]); cephalosporins (ceftiofur [0.25–8 µg/mL; ≥4 µg/mL]); fluoroquinolones (danofloxacin [0.12–1 µg/mL; ≥0.5 µg/mL] and enrofloxacin [0.12–2 µg/mL; ≥0.5 µg/mL]); macrolides (gamithromycin [1–8 µg/mL; ≥8 µg/mL], tildipirosin [1–16 µg/mL; ≥16 µg/mL for P. multocida and ≥8 µg/mL for M. haemolytica], tilmicosin [2–16 µg/mL; ≥16 µg/mL], and tulathromycin [8–64 µg/mL; ≥32 µg/mL]); penicillins (ampicillin [0.25–16 µg/mL; ≥0.06 µg/mL] and penicillin [0.12–8 µg/mL; >0.5 µg/mL]); phenicols (florfenicol [0.25–8 µg/mL; ≥4 µg/mL]); and tetracyclines (tetracycline [0.5–8 µg/mL; ≥4 µg/mL]). Isolates were considered either susceptible or nonsusceptible for statistical analysis based on their MIC and CLSI breakpoints. Nonsusceptible classification was defined as isolates for which antimicrobial drugs’ MIC were above the breakpoint for classification as susceptible, following CLSI guidelines.

Statistical Analysis

Descriptive analysis was performed to describe the distribution of BRD isolates across farms, age groups, and seasons. A descriptive analysis for the distribution of the MIC for P. multocida and M. haemolytica by individual drugs based on CLSI breakpoints was also conducted for all tested antimicrobials in the study. To evaluate which factors (farm, age group, and season) were associated with P. multocida and M. haemolytica isolates nonsusceptible to each of the antimicrobials (each drug tested against each pathogen individually), a logistic regression model was built with a logit link using PROC GLIMMIX in SAS 9.4. The model was as follows:logit[P(y_ijk = 1)] = µ + Age_i + Farm_j + Season_kwhere µ was the baseline log-odds of detecting nonsusceptible isolates (intercept), followed by Age (i = calf, heifer, or cow), Farm (j = 1 to 3), and Season (k = SpringSummer or FallWinter). Considering the uneven distribution of BRD clinical cases culture positive for the isolates across seasons, the lack of more defined seasons as present in some states in North America, and following the results from showing that spring and summer in California carry an increased risk of BRD detection compared with winter, seasons were combined as spring and summer versus fall and winter. Given the lack of sufficient variability across factor levels, such as the uneven distribution of age group and farm isolate susceptibilities across seasons, interactions were not tested. Multiple comparisons were performed using the Tukey honestly significant difference test whenever a significant effect was detected for a variable. The lines option in PROC GLIMMIX was used to summarize group comparisons. Statistical significance was considered when P ≤ 0.05. Further, Pearson chi-squared tests and Cohen’s kappa coefficients (k) were calculated for the susceptibility patterns of co-occurring isolates (i.e., P. multocida and M. haemolytica recovered from the same animal) to identify patterns of antimicrobial susceptibility between the 2 species using PROC FREQ in SAS with the CHISQ and AGREE options. For example, we examined whether a P. multocida isolate was susceptible to ceftiofur when the M. haemolytica isolate from the same animal was also susceptible to ceftiofur.

RESULTS

A total of 301 unique animals were diagnosed with BRD across the 3 evaluated California dairies, 3 animal groups (calves, heifers, and cows), and 2 seasons (clustered into spring and summer vs. fall and winter; over a 2-yr sampling period in the study. A total of 146 unique clinical BRD cases (e.g., each animal present only once in the database) were culture positive for P. multocida, and 63 were culture positive for M. haemolytica, whereas 32 samples showed the co-occurrence of P. multocida and M. haemolytica. The MIC distribution for P. multocida across various antimicrobials is presented in, whereas a more detailed overview of the distribution across seasons, farms, and age groups is presented in it. Regarding overall susceptibility, ceftiofur had the greatest susceptibility among the tested isolates (100%), whereas ampicillin (100%) was the antimicrobial with the highest percentage of nonsusceptible isolates. Because of the lack of variability in susceptibility classification for ampicillin and ceftiofur, these 2 antimicrobials could not be tested in the logistic regression model.

Count distribution of the minimum inhibitory concentration (MIC) for Pasteurella multocida by individual drug based on the Clinical and Laboratory Standard Institute panel. Areas highlighted in blue correspond to susceptible classification, and areas highlighted in red correspond to nonsusceptible (intermediate or resistant) classification. I = intermediate breakpoint classification; MIC = minimum inhibitory concentration; R = resistant breakpoint classification; S = susceptible breakpoint classification.

When examining the descriptive analysis across seasons, farms, and age groups, it was observed that the prevalence of nonsusceptible isolates varied mostly across age groups and farms. Little to no difference was observed for season in terms of susceptibility to antimicrobials in P. multocida isolates by descriptive analysis, and no association was detected in the logistic regression model. A more prominent pattern was observed for the farm effect, in which farm 3 had greater odds of having nonsusceptible isolates than farms 1 and 2 for almost all antimicrobials but penicillin. Age group was also associated with antimicrobial susceptibility of P. multocida isolates, with especially calves having greater odds of harboring nonsusceptible isolates when compared with cows for tetracycline (calf vs. cow: odds ratio [OR] = 12.4, 95% CI: 2.82–54.7) and spectinomycin (calf vs. cow: OR = 6.57, 95% CI: 1.64–26.4).

Table 1. Logistic regression models to test factors associated with the antimicrobial susceptibility of Pasteurella multocida isolates detected in animals diagnosed with bovine respiratory disease

Significant contrasts based on the Tukey adjustment for multiple testing.

For the 32 animals that had growth of both P. multocida and M. haemolytica, their antimicrobial susceptibility tests showed that despite being isolated from the same animal, their susceptibility differed across antimicrobial agents. Except for penicillin, florfenicol, and spectinomycin, the Pearson chi-squared test showed that co-occurring isolates had similar susceptibility patterns to the tested antimicrobials. However, except for ceftiofur, to which all co-occurring isolates were susceptible and had perfect Cohen’s kappa agreement, the degree of the similarity in antimicrobial patterns among co-occurring isolates varied. Gamithromycin (k = 0.67; 95% CI: 0.42–0.92), tilmicosin (k = 0.69; 95% CI: 0.43–0.94), and tildipirosin (k = 0.67; 95% CI: 0.42–0.93) were the antimicrobials to which co-occurring isolates had the greatest agreements. The lowest Cohen’s kappa agreements in antimicrobial susceptibility for co-occurring isolates were for florfenicol (k = −0.10; 95% CI: −0.22–0.01), penicillin (k = −0.06; 95% CI: −0.18–0.06), and spectinomycin (k = 0.08; 95% CI: −0.07–0.23), which were reported earlier in this section to be independent in antimicrobial susceptibility among isolates based on Pearson chi-squared tests. The antimicrobial susceptibility in co-occurring isolates for tetracycline, a drug with a widespread number of nonsusceptible isolates in the study, was low to moderate (k = 0.33; 95% CI: 0.06–0.60).

Table. Co-occurrence analysis between antimicrobial susceptibility of P. multocida (n = 32) and M. haemolytica (n = 32) isolated from 32 animals using their outcome as susceptible or nonsusceptible to a given antimicrobial

The percent distribution and count in parentheses represent values within each category (e.g., none of the co-occurring P. multocida and M. haemolytica are nonsusceptible to ceftiofur).2

Pearson chi-squared test was used to test independence between co-occurring P. multocida and M. haemolytica isolates in their antimicrobial susceptibility per antimicrobial by comparing observed and expected frequencies in a contingency table.

Cohen’s kappa (k) estimates near or equal to 1 indicate high agreement to the susceptibility to a given antimicrobial in co-occurring P. multocida and M. haemolytica isolates; kappa estimates near or below zero indicate a lack of agreement between the 2 isolates to susceptibility results.

In this blog, we cover everything you need to know to respond quickly and confidently when your dog has diarrhea. For further questions or professional guidance, feel free to contact us at professionaldvm129@gmail.com.