Evaluation of the Impact of Antimicrobial Use Protocols in Porcine Reproductive and Respiratory Syndrome Virus-Infected Swine on Phenotypic Antimicrobial Resistance Patterns

Of the 108 pigs placed, 106 (98.1%) completed the trial. Two pigs in the Intensive group were euthanized at 8 days and 13 days post-PRRSV challenge after being assessed with severe clinical signs (score C with the Individual Pig Care system). In total, 1,620 IPC assessments were made of pigs in each of the Minimal and Moderate groups, and 1,580 assessments of pigs in the Intensive group (less due to the mortalities). In the Minimal group (no PRRSV challenge), 98.8% of assessments were scored as normal, compared to 90.0% and 91.6% normal scores for the Moderate and Intensive groups (both PRRSV challenged), respectively. Pigs with mild signs of clinical disease (IPC score A) were identified in 12 (0.7%), 131 (8.1%), and 114 (7.2%) assessments of the Minimal, Moderate, and Intensive groups, respectively. Signs of moderate clinical disease (IPC score B) were recorded in 8 (0.5%), 31 (1.9%) and 18 (1.1%) assessments in the Minimal, Moderate, and Intensive groups, respectively. There were no significant differences in average daily weight gain (ADG) among the treatment groups (P = 0.12). However, ADG was numerically greater (P = 0.05) in the unchallenged pigs (0.93 g per day, SD 0.08) than in the pigs challenged with the PRRSV challenge isolate (0.89 g per day, SD 0.09). These observations were consistent with the occurrence of mild PRRS disease in the exposed pigs.

Of the 324 composite fecal samples cultured, E. coli was isolated from 307 (94.7%) and Enterococcus spp. from 292 (90.1%) samples. There were five Enterococcus species identified, including E. hirae (121 total isolates out of 292: 21% in Minimal, 34% in Moderate, 45% in Intensive group), E. faecium (102 total isolates: 48% Minimal, 29% Moderate, 23% Intensive), E. faecalis (62: 37% Minimal, 32% Moderate, 31% Intensive), E. durans (6 isolates, three each in the Minimal and Intensive groups) and E. villorum (one isolate, Intensive group). At the first sampling after weaning (i.e., before any exposure to antimicrobials), 30 of the 33 E. coli isolates were resistant to at least one antimicrobial, with individual isolates resistant to up to 11 antimicrobials. None of the 36 Enterococcus isolates at that sampling were pansusceptible, with individual isolates resistant to between two and nine antimicrobials. Summary MIC results from all sampling periods for E. coli and Enterococcus are presented in Tables S1 and S2, respectively.

Among all fecal E. coli isolates, resistance was most common to tetracycline, ampicillin, and streptomycin (Fig. 1). No significant differences in prevalence of resistance in E. coli isolates were observed between different sampling dates. The sole significant finding for E. coli isolates was with streptomycin between treatment groups. The odds of resistance to streptomycin were significantly higher (P < 0.0001) in the Moderate group than in the Minimal (OR = 5.45 (CI 2.79–10.68) and Intensive (OR = 5.70 (CI 2.6–12.49,) groups, which did not differ significantly (OR = 0.96 (CI 0.38–2.40, P = 0.99).

Among the Enterococcus isolates, resistance was most common to erythromycin, lincomycin, quinupristin-dalfopristin, tetracycline, and tylosin (Fig. 2). No significant differences were identified between treatment groups. Erythromycin, quinupristin-dalfopristin, and tylosin were the only antimicrobials for which significant differences between sampling dates were found for Enterococcus spp. The odds of resistance to erythromycin were less in week 6 compared to week 1 (OR = 0.165 [CI 0.058–0.468, P = 0.02]), and less in week 7 compared to week 1 (OR = 0.097 [CI 0.028–0.335, P = 007]). The likelihood of quinupristin-dalfopristin resistance was less in week 6 compared to week 1 (OR = 0.143 [CI 0.052–0.395, P = 0.005]), and less in week 7 compared to week 1 (OR = 0.098 [CI 0.029–0.333, P = 0.006]). The odds of resistance to tylosin were less in week 6 compared to week 1 (OR = 0.122 [CI 0.043–0.345, P = 0.002]), and less in week 7 compared to week 1 (OR = 0.097 [CI 0.028–0.335, P = 0.007]). No significant differences in odds of resistance to erythromycin, quinupristin-dalfopristin, or tylosin were found in comparing other sampling events to Week 1.

Among the 307 E. coli isolates, 30 isolates (10.1%) were pansusceptible to the Gram-negative panel of antimicrobials; and 94 isolates (30.6%) were resistant to tetracycline alone. Of the 182 E. coli isolates resistant to more than one antimicrobial, 19 different resistance profiles were observed. These resistance profiles most frequently included resistance to tetracyclines, penicillins (amoxicillin-clavulanic acid and ampicillin), and aminoglycosides (gentamicin and streptomycin) (Fig. 3).

Among the 292 Enterococcus samples, only 9 (3.1%) were pansusceptible to the Gram-positive panel of antimicrobials, and another 33 (11.3%) were resistant to a single antimicrobial, either ciprofloxacin, nitrofurantoin, or lincomycin. Among the 250 Enterococci resistant to more than one antimicrobial, 27 resistance patterns were observed, and 161 (55.1%) of all fecal isolates were resistant to three or more antimicrobials. These resistance patterns most frequently included resistance to tetracyclines, lincosamides, macrolides (erythromycin, tylosin) and streptogramins (Fig. 4). We note that intrinsic resistance to several antimicrobial classes (for example, cephalosporins, lincosamides, aminoglycosides and trimethoprim-sulfamethoxazole) is a recognized feature of enterococci, particularly E. faecium and E. faecalis. Comparison of the resistance patterns across the five Enterococcus species is summarized in Table S4.

The pigs included in the study were selected from a single sow herd in the Midwestern region of the United States that was negative for PRRSV, Influenza A virus, and porcine epidemic diarrhea virus, based on routine clinical and diagnostic monitoring. This source herd was a farrow-to-wean operation housing approximately 5,000 sows. Piglets were weaned at an average of 22.5 days of age. Litters were eligible for inclusion in the study if 1) no cross fostering had been conducted (original litters on sows), 2) the sow had not received an antimicrobial treatment since placement in the farrowing barn, and 3) no piglets in the litter had received antimicrobial treatments. Thirty-six sows of mixed parities that met these criteria were identified. Avoiding low viability piglets, three piglets from each respective litter were selected by convenience and ear tagged. The three pigs from each litter were then randomly assigned across three treatment groups. Thus, a total of 108 weaned pigs (N = 36 per treatment group) comprised the study population.

On the day of weaning, the pigs were commingled in a clean, disinfected, and dried trailer and transported to the research facility. The research facility was a newly constructed barn that had not previously housed any animals prior to placement of the study population. The pigs were weighed on arrival at the research facility then moved into separate rooms by treatment group. In each room, three pigs were placed in each of 12 pens where they remained for the duration of the study. Solid panels were placed between pens to eliminate direct contact and minimize cross-contamination between pens. Each room used negative pressure ventilation with mechanical filters composed of fiberglass media, having a minimum efficiency reported value (MERV) of 16 (Camfil-Farr, Stockholm, Sweden) installed to separately filter all the air entering and exiting each individual room. Additional measures to prevent transmission of microbes between rooms included showering of personnel moving between rooms; an air-purge chamber at the entrances of rooms with pigs; caretakers using a bench to create a transition point to change into room specific boots; and provision of coveralls, gloves, as well as designated equipment and supplies for each room (54).

Pigs in all treatment groups received 2 ml of modified live PRRSV vaccine (Ingelvac PRRS MLV, Boehringer Ingelheim Vetmedica Inc.) 10 days postweaning to reduce severity of clinical illness in the animals later exposed to wild-type PRRSV. One treatment group was not challenged with PRRSV. At 44 days postweaning, the other two groups were inoculated intramuscularly with 2 × 10^3.5 TCID₅₀ of a PRRSV lineage 1 strain 174 field isolate as described previously (33). Clinical assessment of all pigs was conducted by trained personnel using the Individual Pig Care (IPC) scoring system which categorizes pigs according to the perceived severity and duration of clinical signs of any disease they may be experiencing (55, 56). IPC scoring was performed every 3 days for the first 21 days after weaning; three times weekly on Monday, Wednesday, and Friday for 4 weeks after the PRRSV challenge; then once weekly until the end of the trial. Throughout the trial, all individual treatments of pigs were recorded by ear tag number. Mortalities were recorded, including the weights of dead pigs, and all surviving pigs were weighed on the day of marketing (149 days post weaning). All groups were fed the same diets throughout the trial, except for the antimicrobials included in the feed, which varied by treatment group. Pigs were fed eight different rations from postweaning to marketing, which all met or exceeded the National Research Council requirements for swine (57). In line with common industry practices (58), zinc and copper were fed at pharmacological levels in the early nursery rations. Zinc was included at 3000 ppm for the first 11 days on feed, then 2250 ppm from days 12 to 21 on feed. Subsequently zinc was included at 100 ppm to meet nutritional needs. Copper was included at 225 ppm for the first 21 days on feed, then 162 ppm for the remainder of the study.

The three treatment groups were distinguished by protocols (Minimal, Moderate, Intensive) of antimicrobial use (Fig. 5). In the Minimal Use group, which was not challenged with wild-type PRRSV, antimicrobials (penicillin G procaine (33,000 IU/kg once per day for 3 days) or lincomycin HCl (11 mg/kg once per day for 3 days)) were administered by intramuscular injection to individual pigs for therapeutic purposes when indicated. No antimicrobials were administered in the feed or water. This treatment protocol was also followed for both PRRSV challenged (Moderate and Intensive) groups prior to PRRSV challenge at 44 days post weaning.

In both the Moderate and Intensive groups, starting 5 days post PRRSV challenge (49 days postweaning), tilmicosin phosphate (Pulmotil AC [250 mg/ml], Elanco) was administered in the drinking water via a water medication device, for 5 days (approximately 18.3 mg/kg per day). Subsequently, in the Moderate group, ceftiofur crystalline free acid (EXCEDE for Swine, Zoetis) was administered by intramuscular injection (5 mg/kg once) to individual pigs only when necessary for treatment purposes. No antimicrobials were administered in feed to the Moderate group.

In the Intensive group, in addition to tilmicosin in water, at 7 days postchallenge (51 days postweaning) all pigs received an intramuscular injection of 5 mg/kg ceftiofur crystalline free acid (EXCEDE for Swine, Zoetis). Starting 13 days postchallenge (57 days postweaning), a combination of chlortetracycline (400g/ton to deliver 22 mg/kg BW per day) and tiamulin (35g/ton to deliver 2 mg/kg BW per day) was included in the feed of the Intensive group for 14 days per label requirements. No further antimicrobials were used apart from intramuscular injection of ceftiofur crystalline free acid (5 mg/kg) to one individual pig for treatment purposes.

The overall measure of antimicrobial exposure was Used Daily Doses (UDD), calculated as the total number of animal-days of treatment summed across all active ingredients. This was also expressed in relation to the animal-time at risk as Animal Treatment Days per 1000 animal-days at risk (ATD). The antimicrobial use is summarized in Table 1. A UDD of the combination product of chlortetracycline and tiamulin was defined as 1 day of treatment with the combined product.

Composite fecal samples were collected from each pen at the following days postweaning: 9, 44, 49, 57, 65, 71, 79, 104 and 149 (Fig. 5). Composite samples were collected using the EZ Reach Sponge Sampler (World Bioproducts, EZ-10NB-CELL-G). One Sponge Sampler was used per pen targeting all evident dunging areas of the pen floor. Environmental samples from the wall surfaces at approximately five feet above the floor (a height which pigs could not contact) and pen floor surfaces (totaling approximately 12 square feet of surface area) were also collected prior to placement of the pigs, and after pigs were marketed (postsanitation).

Samples were transported on ice to an American Association of Veterinary Laboratory Diagnosticians accredited veterinary diagnostic laboratory where they were processed within 24 h after delivery. Formal Standardized Operating Procedures of the laboratory were followed for culture and identification of bacteria, and antimicrobial susceptibility testing. Sponge samples were aseptically placed into sterile plastic bags, mixed with buffered peptone water, and stomached for 2 min at 265 rpm. This rinsate was mixed with double strength MacConkey broth or double strength Enterococcosel broth at a 1:1 ratio (BD Difco manufactured by Becton, Dickinson & Co. Sparks, MD) and incubated at 35°C and 45°C for 24 h for E. coli and Enterococcus, respectively. After incubation, a loopful was streaked on Enterococcosel and MacConkey agar (BD Difco manufactured by Becton, Dickinson & Co. Sparks, MD) for Enterococcus spp. and E. coli, respectively, then incubated at 35°C for 24 h. For samples with characteristic colonies present on the selective agar plates, one typical well-isolated colony was streaked to blood agar and incubated overnight at 35°C. One colony from the blood agar plates was used for bacterial identity confirmation using Matrix Assisted Laser Desorption Ionization – Time of Flight (MALDI-TOF) technology.

Antimicrobial susceptibility testing was performed using the TREK broth microdilution system (Trek Diagnostic Systems, Cleveland, OH). Enterococcus spp. samples were tested with the Sensititre^TM Gram Positive NARMS antimicrobial plate (16 antimicrobials plus positive and negative controls). E. coli samples were tested with the Sensititre Gram Negative NARMS antimicrobial plate (14 antimicrobials plus positive and negative controls). Susceptibility breakpoints were adopted from Clinical and Laboratory Standards Institute (CLSI) standards according to NARMS Human Isolate reports (59) as summarized in Table 2. The susceptibility results were dichotomized into resistant or susceptible according to the MIC breakpoints, with intermediate responses being categorized as susceptible. Because CLSI breakpoints have not been established for E. coli for streptomycin or azithromycin, NARMS-established interpretive standards were used as the breakpoints for these antimicrobials (59). Susceptibility results were further analyzed by antimicrobial class.

For most antimicrobials, only descriptive analysis was reported due to the sparsity of data, i.e., low prevalence of resistance. To analyze potential associations between AMU and AMR in E. coli and Enterococcus spp., a generalized estimating equation (GEE) model was generated using SAS software, Version 9 (SAS Institute Inc., Cary, NC, USA). Model effects included treatment group and sample date, alone or in combination, as well as an exchangeable working correlation structure (repeated measure) of pen location within room. Additionally, a binomial distribution and logit function were specified to reflect the dichotomous nature of the outcome variable (i.e., susceptible/resistant). The primary outcome was the antimicrobial susceptibility status of each isolate to the antimicrobial of interest. In initial model evaluations, only three model including a treatment group by sample-date interaction term converged and the Quasi Information Criterion (QIC) indicated that these were not the preferred models. The QIC criteria indicated that the most parsimonious models were those that did not include sample date for most bacteria-antimicrobial combinations. However, as analysis of the temporal effects of the antimicrobial exposure was a primary goal of the study, the final model selected included both treatment group and sample date as well as the repeated measures factor. Model comparison data are presented in Table S5. For antimicrobials for which resistance was relatively common and the model converged, odds ratio estimates were calculated with 95% confidence intervals for sample date (including erythromycin, quinupristin-dalfopristin, and tylosin for Enterococcus isolates) and treatment group (including streptomycin for E. coli) comparisons. Statistical significance was indicated at P < 0.05 after adjusting for multiple comparisons using the Tukey-Kramer method. Prevalence of resistant isolates was compared between treatments, with statistical significance indicated at P < 0.05. One-way ANOVA was used to compare average daily gain between treatment groups and a two-sample t test was conducted to compare average daily gain between PRRSV challenged versus nonchallenged groups (Statistix 10.0. Analytical Software, Tallahassee, FL, USA).