Effects of antimicrobials fed as dietary growth promoters on faecal shedding of Campylobacter, Salmonella and shiga-toxin producing Escherichia coli in swine


  • Mention of a trade name, proprietary product or specific equipment does not constitute a guarantee or warranty by the USDA and does not imply approval to the exclusion of other products that may be suitable.

  • USDA is an equal opportunity provider and employer.


James E. Wells, USDA, ARS, US Meat Animal Research Center, PO Box 166, Clay Center, NE 68933, USA. E-mail: jim.wells@ars.usda.gov



To determine whether antimicrobials commonly used in swine diets affect zoonotic pathogen shedding in faeces.

Methods and Results

Barrows (n = 160) were sorted into two treatments at 10 weeks of age (week 0 of the study), and fed growing, grow finishing and finishing diets in 4-week feeding periods. For each feeding phase, diets were prepared without (A−) and with (A+) dietary antimicrobials (chlortetracycline, 0–8 week; bacitracin, 9–12 week) typical of the United States. At week 0, 4, 8, 9, 10 and 12 of the study, faecal swabs or grabs were collected for analyses. Campylobacter spp. was absent at week 0, but prevalence increased over time with most isolates being identified as Campylobacter coli. When chlortetracycline was used in A+ diets (week 4 and 8), prevalence for Campylobacter spp., pathogenic Escherichia coli O26 and stx genes was lower in faeces. On week 12 after the shift to bacitracin, Campylobacter spp. and stx genes were higher in faeces from piglets fed A+ diet. Pathogenic E. coli serogroups O103 and O145 were isolated throughout the study and their prevalence did not differ due to diet. Pathogenic E. coli serogroups O111 and O121 were never found in the piglets, and Salmonella spp. prevalence was low.


In production swine, growing diets with chlortetracycline may have reduced pathogen shedding compared with the A-growing diets, whereas finishing diets with bacitracin may have increased pathogen shedding compared with the A-finishing diet.

Significance and Impact of the Study

Inclusion of antimicrobials in the diet can affect zoonotic pathogen shedding in faeces of swine.


Salmonella spp. and Campylobacter spp. are foodborne pathogens commonly associated with swine production. A number of research studies have examined the prevalence of these pathogens in swine production facilities, but few studies have examined the role of dietary antimicrobials and management in pathogen persistence through a production cycle (Fosse et al. 2009). In a survey of Salmonella spp. prevalence on swine farms, Gebreyes et al. (2006) observed that U.S. producers utilizing antimicrobial-free diets had higher prevalence compared with conventional producers utilizing antimicrobials, whereas in a separate survey in France (Rossel et al. 2006), the addition of antimicrobials to diets was associated with higher odds ratio for finding Salmonella-seropositive pigs at slaughter. Neither report noted specific dietary antimicrobials associated with differences in Salmonella prevalence. In previous research with swine fed diets with and without dry skim milk, we noted that Salmonella spp. and Campylobacter spp. were low with subtherapeutic dietary chlortetracycline, whereas these pathogens appeared to be higher in the pigs after the dietary antimicrobial was switched to bacitracin prior to slaughter (Wells et al. 2005). In addition, levels of Lactobacillus spp. and total Escherichia coli counts appeared to be reduced with the dietary chlortetracycline and appeared to increase when diets were switched to bacitracin, but the study lacked a control diet formulated without antimicrobial for comparison to determine whether dietary antimicrobials alter pathogen shedding and indigenous flora.

Strains of shiga-toxin-producing E. coli (STEC) are serious foodborne pathogens that can result in human illness and possibly death. Escherichia coli O157:H7 is the most widely recognized STEC associated with human disease, but this pathogen has rarely been reported in pork products (Ferens and Hovde 2011). Non-O157 STECs include a variety of serotypes, but O26, O45, O103, O111, O121 and O145 have been associated with human illnesses and may account for 70% of the non-O157:H7 STEC human disease cases in the United States (Brooks et al. 2005). The pathogenic non-O157 STEC appears to be transmitted to humans via similar vehicles and can cause similar forms of illnesses, mainly self-limiting to severe diarrhoea, and a potentially life-threatening haemolytic uraemic syndrome as with E. coli O157:H7. Little is known about the potential distribution, if any, of non-O157 STEC in swine.

To determine how dietary antibiotics as growth promoters may influence pathogen shedding, we designed a study targeting the growing and finishing phases of swine production. One-half of the animals were fed diets with antibiotics and one-half were fed diets without antibiotics. Faecal samples were collected at different phases of the swine production cycle and enriched for Salmonella spp., Campylobacter spp., stx genes and non-O157 STEC.

Materials and methods

Animals, design and management

The experimental protocol was reviewed and approved by the Animal Care and Use Committee of the US Meat Animal Research Center (USMARC). A total of 160 barrows from the USMARC swine population (composite of Landrace, Yorkshire and Duroc) born in the same farrowing period (<1-week period) were used in this study conducted at the USMARC swine facilities. Pigs in the USMARC swine nursery were acclimated to the growing ration without antimicrobials at 7 weeks of age (3 weeks prior to the study). At approximately 8 weeks of age (2 weeks prior to the study), animals were weighed for treatment assignments. Animals were blocked by weight, and litter, and randomly sorted into one of the two dietary treatment groups. Animals were housed in 20 pen groups of eight animals per pen, with ten pens (80 animals) phase-fed diets with antimicrobials as growth promoters and ten pens (80 animals) fed the same diet without the antimicrobials.

At approximately 10 weeks of age (week 0 of the study), animals were weighed and transferred into respective pens in USMARC growing/finishing barn for the study (designated day 0 for study purposes). Diets (Table 1) were formulated to meet or exceed NRC suggested dietary recommendations (NRC 1998). Standard USMARC growing/finishing swine dietary regime was utilized and the regime follows typical swine production practices (Yen et al. 2004). The USMARC swine growing ration was fed from weeks 10 to 13 of age (up to week 4 of the study), the USMARC swine growing/finishing ration was fed from weeks 14 to 17 of age (up to week 8 of the study), and the USMARC swine finishing ration was fed from week 18 of age until slaughter (up to week 12 of the study). The growing and grow/finishing rations were formulated with and without chlortetracycline (0 or 0·1% of diet as premix) and finishing ration was formulated with and without bacitracin (0 or 0·025% of diet as premix). Feed intake was monitored in five pens per treatment by weighing feed provided and feed remaining on each animal weigh date. Animals were weighed at initiation of the study and every 4 weeks until completion of the study.

Table 1. Composition of basal diets for each phase of the study (as-fed basis)
Time of studyWeeks 0–4Weeks 5–8Weeks 9–12
Production phaseGrowingGrow-finishingFinishing
  1. a

    Supplied 55 ppm chlortetracycline in growing and growing-finishing diets, and 33 ppm bacitracin methylene disalicylate in finishing diets.

  2. b

    Based on chemical composition of feed ingredients published by NRC (1998).

Ingredient, %
Soybean meal 48% CP19·0013·359·58
Soybean oil1·090·921·00
Dicalcium phosphate1·000·880·67
Vitamin premix0·200·200·20
Trace mineral premix0·200·200·20
Choline chloride0·100·100·10
Antibiotic (or Corn)a0·100·100·025
Calculated compositionb
Digestible energy, MCal kg−13·503·503·50
Crude protein, %15·6513·4311·93
Ca, %0·620·500·45
P, %0·530·490·43

Sample collection

At approximately 10 weeks of age while still in the swine nursery (representing week 0 of the study), rectal swabs (two each animal each sampling) were taken from the piglets designated for trial for determinations of pathogen prevalence in the samples, and faecal grab samples were taken from two of eight piglets of each designated pen for future commensal bacterial population analysis. Thereafter, on 4-week intervals (week 4, 8 and 12) corresponding to changes in the phase feeding regime, rectal swab samples on all animals and faecal grab samples on ¼ of the animals in each pen were collected. Additional rectal swab samples were collected from all animals on weeks 9 and 10 to evaluate the effect of antimicrobial shift from chlortetracycline in growing diets to bacitracin in finishing diets. Each swab was placed in a separate 15-ml conical centrifuge tube, and faecal grabs were inverted in the glove and placed in a clean plastic bag. Sample processing was done as described below to monitor the level and prevalence of each pathogen in faeces for each of the treatment groups until slaughter.

Skin swab samples were collected prior to slaughter to determine the level and prevalence of each pathogen on skin. The skin was sampled using a premoistened sponge with 20 ml of buffered peptone water. Approximately ten side-to side strokes on one side of each pig was used for sampling an area of approximately 1500 cm2 for the skin samples.

Microbiological analyses

Samples were processed to enumerate the level of Salmonella on faecal and skin swabs as described previously (Britcha-Harhay et al. 2008). For the faecal swabs, a 2-ml volume of tryptic soy broth (TSB; Becton, Dickinson and Company, Spark, MD, USA) was added to each conical tube, mixed by vortexing and 300 μl removed to a 2-ml cluster tube. For the skin samples, 500 μl was transferred to a 2-ml cluster tube. For each sample, 50 μl from the cluster tube was spiral plated onto XLDtnc plates prepared with XLD agar (Becton, Dickinson and Company) containing tergitol (4·6 ml l−1), novobiocin (15 mg l−1) and cefsulodin (10 mg l−1). Plates were incubated at 37°C for 18–22 h and then at room temperature for another 18–22 h. Presumptive colonies were counted and confirmed as Salmonella by invA PCR (Ziemer and Steadham 2003).

Samples were processed to determine prevalence of Salmonella. For faecal swabs, 8 ml of TSB was added to each tube. For skin swabs, 80 ml of TSB was added to each sample bag. Each sample was incubated at 25°C for 2 h and then at 42°C for 6 h prior to being held at 4°C overnight. A 1-ml volume of sample was transferred into a tube containing 13 ml of tetrathionate broth (Difco) and enriched for 48 h at 37°C. A 20-μl volume of the tetrathionate enrichment was transferred to a tube with 10 ml of Rappaport-Vassiliadis Soya (RVS; Oxoid) broth and selectively enriched for Salmonella by incubation for 24 h at 42°C. An enriched sample (20 μl) from RVS broth was streaked onto XLDtnc and Difco Brilliant Green Agar (Becton, Dickinson and Company) to isolate colonies of Salmonella. Salmonella isolates were confirmed by invA PCR (Ziemer and Steadham 2003).

To determine the prevalence of STEC and distribution of stx gene(s) from faecal swabs, 1 ml of TSB enrichment from Salmonella method above was concentrated using immunomagnetic separation (Dynabeads anti-Escherichia coli O26, anti-E. coli O103, anti-E. coli O111 and anti-E. coli O145; Invitrogen Corp., Carlsbad, CA, USA). The recovered beads were suspended into 100 μl of sterile phosphate-buffered saline (PBS) and 20 μl was spread onto USMARC chromogenic medium (Kalchayanand et al. 2012) and incubated overnight at 37°C. This chromogenic medium is based on β-d-galactosidase activity, utilizing a mixture of carbohydrate sources and selective compounds, which allow colour-based separation of these serotypes on a single plate. The colony colours developed on this media are bright turquoise, pale green, dark bluish green and purple for serogroups O26, O103, O111 and O145, respectively. Representative colonies (up to eight colonies) were picked, tested by dry spot (Oxoid) and confirmed by multiplex PCRs for serogroup (Kalchayanand et al. 2012). Enrichments were assayed separately for stx genes and isolates for enterohemorrhagic E. coli (EHEC) virulence gene profiles for eae, hlyA, stx1 and stx2 using multiplex PCR (Paton and Paton 1998).

To determine the prevalence of Campylobacter from faecal swabs, 13-ml volume of Bolton broth (Oxoid) with lysed horse blood and 48 μl Bolton broth selective supplement (Oxoid) were added to each conical tube. For skin swab enrichments, a 1-ml volume from each skin swab was inoculated into Bolton broth as described above. The tubes were sealed and incubated at 37°C for 4 h followed by 42°C for 44 h. A 50-μl volume was streaked onto a Campy-Cefex plate (Stern et al. 1992). Plates were incubated at 42°C for 48 h in microaerophilic environment boxes (Mitsubishi Gas and Chemical, New York, NY, USA) with two microaerophilic sachets per box. Campylobacter were confirmed and species determined using 23s rRNA gene (Eyers et al. 1993; Fermér and Olsson Engvall 1999) and lpxA (Klena et al. 2004) multiplex assays.

Faecal samples were collected from two animals from each pen (20 per treatment, 40 per sample day), diluted in TSB up to 10−5 dilution and spiral plated onto Chromagar ECC (two dilutions per sample; DRG International, Mountainside, NJ, USA) and modified lactobacilli deMan-Rogosa-Sharpe agar (mMRS, Wells et al. 2005; two dilutions per sample). The Chromagar ECC plates were incubated at 37°C overnight. The mMRS agar plates were placed into large anaerobic boxes (with three Anaero-pak sachets per box; Mitsubishi Gas and Chemical) and incubated overnight at 39°C. Colony forming units were counted and recorded for total faecal enumeration calculations.

Histology, IgA and using analyses

After weeks 8 and 12 of the study, eight pigs from each treatment were randomly selected and slaughtered. From each animal, 3- and 10-cm segments of terminal ileum were collected. The 3-cm segments were processed, embedded and stained according to previously described procedures (Luna 1968). Briefly, freshly cut intestinal sections were rinsed in cold PBS and then fixed in freshly prepared chilled fixative solution (FEA: 10 ml formalin, 70 ml 95% ethanol, 15 ml distilled water, 5 ml acetic acid). Intestinal segments were dehydrated over a 2-day period using increasing concentrations of ethanol and chloroform. Sections were embedded in paraffin, and cross-sections were cut approximately 10 μm thick. Sections were stained with hematoxylin and eosin, and morphometric measurements were performed by one person using light microscopy with a computer-assisted morphometric system (Bioquant Image Analysis Corp., Nashville, TN, USA). The height and crypt depth of eight well-oriented villi per sample were measured.

The 10-cm segments collected at slaughter were washed in saline and stored on ice in Gibco Leibovitz's L15 with glutamine medium (Invitrogen, Grand Island, NY, USA). The mucosa and submucosa were carefully removed from the underlying muscle layer, washed in prepared physiological extracellular fluid solution (PEF; Green et al. 2003) warmed to 39°C and aerated with 95% O2/5% CO2 and mounted onto Ussing flux chambers in duplicate per animal. The flux chambers were mounted onto the Ussing apparatus, and each chamber side (luminal and serosal) was bathed with circulating PEF solution (10 ml), maintained at 39°C and bubbled with 95% O2/5% CO2 (Lyte 2004). Bacterial cultures from the USMARC strain collection of Salmonella enterica Typhimurium strain USMARC#622-1 (multidrug resistant; AmApCSSuTe) and Salm. enterica Anatum strain USMARC#647-2 (pansusceptible) by growing overnight in TSB at 37°C and allowed to set at room temperature for 24 h prior to use. On the day of study, Ussing inoculates were prepared by diluting 10 μl of each culture in 10 ml PEF solution. The luminal side of the Ussing chamber was drained, inoculated PEF solution was added and tissues were exposed for 60 min. Tissues were voltage clamped (World Precision Instruments, Sarasota, FL) to monitor tissue integrity and viability (Lyte 2004). At the end of the exposure time, the tissue was removed and gently washed with sterile PBS, pH 7·4 (PBS; Sigma, St Louis, MO). Tissues were cut in one-half on a sterile surface and each one-half weighed in a sterile weigh boat. To determine total bacteria on the tissue, one-half of each tissue sample was washed three additional times in sterile PBS, and to determine internalized bacteria, the other one-half was washed two additional times with sterile PBS and one time sterile PBS with 100 μg ml−1 gentamycin (Sigma) for 80 min. Each tissue half sample was gently blotted dry on a sterile paper towel and transferred to a glass tube with 2 ml PBS. Tissues were macerated (Model 5000; Omni International, Waterbury, CT, USA) and each sample was diluted 101–103-fold in buffered peptone water (Becton, Dickinson and Company). From dilutions of each prepared tissue half, 50 μl were plated onto Difco Hektoen enteric agar (Becton, Dickinson and Company) plates with 0 (for total Salmonella counts) and 100 μg ml−1 ampicillin (Sigma; for Salm. enterica Typhimurium counts). Plates were incubated overnight at 37°C, and colonies were counted. Counts for S. Anatum were calculated by difference in the two plates with and without ampicillin. Per cent total Salmonella bound was calculated as a ratio of CFU ml−1 in the macerated tissue sample (normalized to 10 ml volume) to CFU ml−1 inoculated into the Ussing chamber. Per cent Salmonella internalized was calculated as a ratio of CFU ml−1 in the macerated tissue with gentamycin to CFU ml−1 in macerated tissue without gentamycin.

At week 0 and after weeks 4, 8 and 12 of the study, blood was collected via jugular venipuncture and placed immediately on ice. After collection, blood samples were centrifuged at 800 g for 10 min, with plasma collected and frozen at −20°C until further analyses. Plasma was analysed for immunoglobulin A (IgA) using a commercial enzyme-linked immunosorbent assay performed in 96-well plates (Bethyl Laboratories, Inc., Montgomery, TX, USA).

Statistical analyses

Results from the enumerations were log transformed, and these results were analysed by anova using general linear models (GLM) procedures in SAS (SAS Institute, Cary, NC, USA) with main effect of dietary treatment and pen as the experimental unit. Effects of diet on gain and intestinal tissue measures were analysed using GLM with individual animal as the experimental unit and initial weight (for gain) or weight (for intestinal tissue measures) as a covariant (Oliver and Miles 2010; Oliver et al. 2012). Effects of diet on intake and gain-to-feed ratios were analysed using GLM with pen as the experimental unit and initial pen weight as a covariant. Pathogen prevalence with treatments within each trial was analysed individually against the basal diet and other treatments using the Fisher Exact probability test and the P-value reported is for the same or stronger association (Uitenbroek 1997). For all statistical analyses, differences were considered significant when P-values were <0·05 and tendencies between 0·05 and 0·1.


Initial weights for barrows were similar across treatment groups (33·45 and 33·35 kg of body weight for A− and A+, respectively; > 0·10) prior to dietary treatment. Average body weights in kg were 58·15 and 58·12 on week 4 of the study (= 0·43, 14 weeks of age), 83·41 and 85·33 on week 8 (= 0·09; 18 weeks of age) and 102·18 and 102·02 on week 12 (= 0·87; 22 weeks of age), for A− and A+, respectively. The average daily gains in kg were 0·85 and 0·85 over weeks 0–4 (= 0·60), 0·94 and 0·99 over weeks 4–8 (= 0·02) and 0·88 and 0·88 over weeks 8–12 (= 0·80), for A− and A+ treatments, respectively. In the five pens monitored for feed intake, feed intakes in kg were 10·23 and 11·14 over weeks 0–4 (= 0·16), 16·30 and 17·07 over weeks 4–8 (= 0·14) and 23·91 and 23·70 over weeks 8–12 (= 0·89) for A− and A+ treatments, respectively. Gain-to-feed ratios were 0·49 and 0·47 over weeks 0–4 (= 0·49), 0·42 and 0·40 over weeks 4–8 (= 0·46) and 0·30 and 0·32 over weeks 8–12 (= 0·40) for A− and A+ treatments, respectively. Intestinal tissue morphologies were determined at the end of the study for random pigs of similar body weight. Pigs fed the A− diet exhibited similar villi height and crypt depth compared with pigs fed the A+ diet (> 0·10; Fig. 1).

Figure 1.

Histology results for jejunum and ileum tissues collected from pigs fed diets without (A−) and with (A+) dietary antimicrobials for 12 weeks (error bars represent SEM). (■) A+ and (image_n/jam12065-gra-0001.png) A−.

Faecal counts for lactic acid bacteria (LAB) were similar for both treatment groups prior to dietary treatment (8·20 and 8·23 log10 CFU g−1 faeces for A− and A+, respectively). Faecal counts for LAB increased slightly from week 0 to 8 for both treatments (Fig. 2), and by week 12, faecal LAB counts for A− and A+ were 8·66 and 8·41, respectively (= 0·14). Total Escherichia coli counts were similar for both treatment groups prior to dietary treatment (6·26 and 6·34 log10 CFU g−1 faeces for A− and A+, respectively; > 0·10), and counts decreased from week 0 to 8 (5·03 and 5·08 log10 CFU g−1 faeces for A− and A+, respectively). When measured on week 12 (5·81 and 5·93 log10 CFU g−1 faeces for A− and A+, respectively), the total E. coli counts increased compared with week 8 (< 0·001), but there was no difference between treatments (= 0·65).

Figure 2.

Bacterial counts in faeces for lactic acid bacteria (LAB; image_n/jam12065-gra-0002.png, image_n/jam12065-gra-0003.png) and Escherichia coli (E. coli; image_n/jam12065-gra-0004.png, image_n/jam12065-gra-0005.png) from pigs fed diets without (A−; open symbols) and with (A+; closed symbols) dietary antimicrobials for 12 weeks. Antimicrobial supplemented to diet in weeks 0 to 8 was chlortetracycline and in weeks 9 to 12 was bacitracin.

The prevalence for Campylobacter spp. was 0% for both treatments at the initiation of the study and increased for all pigs to 5% on week 4, 15% on week 8 and 28% on week 12. Overall for the entire 12-week study, treatment groups did not differ in the faecal prevalence for Campylobacter, but there were significant interactions with time. Faecal prevalence for Campylobacter was 6·25 and 3·75% on week 4 and 20·0 and 10·0% on week 8 for A− and A+ diets, respectively (Fig. 3). Cumulatively during the growing and growing-finishing phases when chlortetracycline was used in the A+ diets, faecal prevalence was 13·1 and 6·9% for A− and A+ diets, respectively (= 0·04). Faecal prevalence changed little from week 8 to 10 for either treatment, but by week 12 at the end of the finishing phase when bacitracin was used in the A+ diet, faecal prevalence was 18·8 and 38% for A− and A+, respectively (= 0·007). Within treatments, a significant increase in the prevalence was noted for A+ diet over time from weeks 10 to 12 (= 0·023), but Campylobacter spp. were not different within the A− group over the same time period. All Campylobacter spp. were confirmed as thermophilic Campylobacter, and 93·5% of these isolates were specifically identified as Campylobacter coli. Salmonella was only detected in samples at the start of the study and on week 4, but prevalence was low (<2·5% each sample time) and no treatment effects were detected.

Figure 3.

Prevalence for Campylobacter spp. in faeces from pigs fed diets without (A−; image_n/jam12065-gra-0002.png) and with (A+; image_n/jam12065-gra-0003.png) dietary antimicrobials for 12 weeks. Antimicrobial supplemented to diet in weeks 0 to 8 was chlortetracycline and in weeks 9 to12 was bacitracin.

Enriched faecal swab samples were assayed for the presence or absence of stx (stx1 and stx2) after the initiation of the study and on weeks 4, 8, 9, 10 and 12, and the overall percentage of samples stx-positive samples were 21·9, 20·0, 32·5, 58·1 and 58·0%, respectively. Enriched samples were not assayed on week 0. Overall, A− and A+ did not differ in the number of enriched faecal samples found positive for stx, but there was a significant interaction with time. The percentages of enriched samples positive of stx were 26·25 and 17·5% on week 4 and 23·75 and 16·25% on week 8 for A− and A+ diets, respectively (Fig. 4). Cumulatively for the growing and growing-finishing phases when chlortetracycline was used in the A+ diets, the percentages of enriched faecal samples positive for stx were 25·0 and 16·88% for A− and A+ diets, respectively (= 0·05). The percentage of samples positive for stx increased similarly for both treatments for week 9, but by weeks 10 and 12 at the end of the finishing phase, the presence of stx increased more in the A+ treatment when bacitracin was used in the diet with more than 65% of the samples testing positive for stx1 and stx2 (< 0·01). In the faecal samples tested for the presence of stx, 97·4, 1·3 and 1·3% were positive for only stx2, only stx1, and both stx1 and stx2, respectively.

Figure 4.

Prevalence for stx genes in faeces from pigs fed diets without (A−; image_n/jam12065-gra-0002.png) and with (A+; image_n/jam12065-gra-0003.png) dietary antimicrobials for 12 weeks. Antimicrobial supplemented to diet in weeks 0 to 8 was chlortetracycline and in weeks 9 to 12 was bacitracin.

Enriched samples were analysed for select non-O157 STEC serogroups, and isolates of E. coli O26, O103, O121 and O145 with virulence genes common to EHEC and STEC were recovered from faecal swab samples collected at weeks 4, 8, 10 and 12 of the study. Samples collected on week 0 were not analysed. On week 4 of the study, the prevalence for A− and A+ was 22·5 and 2·5%, respectively, for E. coli O26 isolates testing positive for virulence genes (= 0·0084), but no other significant differences between A− and A+ diets were noted in regard to the presence or absence for the other EHEC/STEC isolates on week 4 and for all EHEC/STEC isolates on weeks 8, 10 and 12 (> 0·05; data not shown). For pooled data for both treatments, potentially pathogenic E. coli (EHEC or STEC) O26, O103 and O145 were isolated from 6·9, 2·4 and 4·8%, respectively, for all of the samples tested (Table 2). Samples were negative for E. coli O111 isolates, and none of the O121 isolates tested positive for EHEC/STEC virulence genes. Over the course of the study, EHEC/STEC O26, 0103 and O145 were cumulatively 35·0, 10·0, 3·1 and 0·6% of samples on week 4, 8, 10 and 12, respectively (< 0·05).

Table 2. Detection of major non-O157 enterohemorrhagic Escherichia coli (EHEC) or shiga-toxin-producing E. coli (STEC) E. coli serogroups associated with human illness in enrichments from swine faecal samples collected for growing (week 4; 14 weeks of age), grow-finishing (week 8; 18 week of age) and finishing (weeks 10 and 12; 20 and 22 weeks of age, respectively) phases of swine production
Week of studyEscherichia coli serogroups from enrichments
Serogroup positive (%)/Serogroup positive with virulence gene(s) (%)a
  1. a

    Values to the left of the bar represent percentage of samples with isolates positive for the respective serogroup and the values to the right of the bar represents percentage of samples with isolates positive for the respective serogroup and virulence gene(s) (eae, stx1, stx2 or hlyA).

Week 432·5/25·00·0/0·00·0/0·00·0/0·017·5/17·540·0/35·0
Week 85·0/2·510·0/7·50·0/0·00·0/0·00·0/0·015·0/10·0
Week 100·0/0·03·1/1·90·0/0·00·0/0·01·2/1·24·3/3·1
Week 120·6/0·00·0/0·00·0/0·07·5/0·01·2/0·69·3/0·6

Overall, 65·3% of the isolates for E. coli O26, O103 and O145 were positive for virulence genes. A breakdown of the serotype and virulence genes is presented in Table 3, and 40·6% of these isolates were E. coli O26 isolates with eae only. Escherichia coli O145 and O103 isolates with stx1, eae and hlyA represented 34·4 and 15·6% of the isolates, respectively. Escherichia coli O26 isolates with stx2, eae and hlyA, E. coli O103 isolates with stx1, and E. coli O145 isolates with stx1 and eae each represented 3·1% of the total isolates.

Table 3. Distribution of potentially pathogenic Escherichia coli isolates as determined by serogroups and virulence genes
E. coli serogroupsVirulence gene combination
eae onlystx onlystx, eaestx, eae, hlyA
Per cent positives (% all isolates with virulence genes)

On the skin swabs collected from each pig at the end of the study (week 12), Campylobacter spp. was found in 26·3 and 50·0% of the A− and A+ treatment samples, respectively (= 0·003). Salmonella spp. and stx genes were not found in enriched skin swab samples. Escherichia coli O26 testing positive for eae and hlyA were found in skin swab enrichments from three samples from the A+ treatment, but there was no significant treatment effect and none of the other serotypes were found.

Serum IgA concentrations at the beginning of the study were 121·7 and 114·9 μg ml−1 for animals sorted into treatments A− and A+, respectively. On week 4, plasma IgA concentrations were similar for both treatments (172·2 and 160·9 μg ml−1 for A− and A+, respectively). On weeks 8 and 12, plasma IgA concentrations were higher for animals fed A− diets compared with animals fed A+ diets (239·1 and 119·8 μg ml−1 vs 173·1 and 152·8 μg ml−1, respectively; < 0·05). Attachment of Salmonella strains to terminal ileum mucosa was measured with inoculated Ussing chambers and did not differ between growing-finishing and finishing phases within each diet. When combined across the phases and compared (Table 4), A− and A+ diets did not differ in Salmonella attachment to mucosa (3·2 and 2·9%; = 0·74). With mucosa from the animals fed A− diet, 1·4% of the attached Salmonella spp. were internalized, and from animals fed A+ diet, 4·5% were internalized (= 0·078). Specific internalization into mucosa of multidrug resistant (MDR) Salmonella enterica Typhimurium using ampicillin selection was 36·4 and 47·8% of the recovered isolates for A− and A+ diets, respectively (= 0·1). Results from the Ussing chamber studies also were analysed separately with either villi height or crypt depth as covariates in the statistical model, and the covariates did not change the interpretation of the data except in the case of specific internalization into mucosa for MDR Salm. enterica Typhimurium, where the P-value changed from = 0·10 to = 0·35 when crypt depth was used in the model.

Table 4. Attachment and internalization of Salmonella enterica with terminal ileum mucosal samples collected from animals fed diets without (A−) and with (A+) dietary antimicrobials during growing-finishing (±chlortetracycline) and finishing (±bacitracin) phases of production
ParameterDietary treatmentSEM P
  1. a

    Total for both Salm. enterica Anatum and Typhimurium inoculated.

  2. b

    Represents Salm. enterica Typhimurium internalized.

Per cent total Salmonella bound2·93·20·50·74
Per cent bound Salmonella internalized (% total bound)a4·51·40·80·08
Per cent Ampr Salmonella internalized (% total internalized)b36·447·84·40·10


Dietary antimicrobials are used to promote growth and prevent disease in food animals (Cromwell 2001), but in recent years, the use of these management tools has come under scrutiny due to the potential risk for generating antibiotic-resistant bacteria that may impact human health (Phillips et al. 2004). In food-producing animals, tetracyclines (including chlortetracycline) are the most commonly used drug class in the United States, and bacitracin is in the polypeptide class that is not reported separately (Food and Drug Adminstration 2009). In the US swine production, chlortetracycline is the most commonly used antimicrobial through all phases of production, and bacitracin is the most commonly fed antimicrobial in the later phases of production, based on recent farm surveys (Apley et al. 2012). In the current study, we did not see a significant benefit in growth or performance attributable to subtherapeutic dietary antimicrobials and did detect expected levels of antimicrobials in faeces from each respective treatment group (Varel et al. 2012). We observed a tendency for better gain at the end of the grow-finishing phase after 8 weeks of dietary chlortetracycline in the antimicrobial-fed swine, but this effect diminished after the switch to bacitracin as measured at 12 weeks. In young piglets, dietary antimicrobials altered gastrointestinal mucosa with respect to villi height or crypt depth of the jejunum and ileum (May et al. 2012), but these measures did not differ in the current study with older pigs when feeding antimicrobials.

Dietary antimicrobials also may prevent disease and reduce harmful pathogens. A recent study by Hurd et al. (2008) with swine at harvest concluded that there was a correlation between animal health and human health, thereby concluding that animal management decisions on the farm were expected to directly impact public health. Infection by Campylobacter spp. is one of the most observed bacterial illnesses in humans in the United States. (Scallan et al. 2011). In swine samples from this study, we observed Campylobacter spp. shedding overall in 12·1% of the faecal samples, with pathogen shedding lowest at the beginning of the study (0%) and highest prior to harvest (28·1%). Previously, we had observed negligible shedding of Campylobacter spp. in nursery swine as a consequence of carbadox and copper sulfate antimicrobials in their diet (Wells et al. 2010), and our low prevalence on week 0 is likely a consequence of the prior nursery diet. On farms in Ontario, approximately 35% of finishing pigs were positive for Campylobacter spp., and was the swine production age group exhibiting the highest faecal prevalence (Farzan et al. 2009). Similar to our study, almost all of the Campylobacter spp. could be identified as Campylobacter coli.

Hurd et al. (2008) observed for a US study, a two- to four-fold increase for per cent Campylobacter spp. positive swab samples collected at harvest for pigs fed conventional diets with antibiotics compared with those fed antibiotic-free diets. In contrast, Tadesse et al. (2011) and Thakur and Gebreyes (2005) observed a similar prevalence of c. 50% for Campylobacter spp. isolated from faeces from different production systems with finishing pigs and did not see any effect of antibiotics in the diet. However, none of these studies reported the type of antimicrobial being fed in the conventional systems monitored. Compared to the pigs on the antibiotic-free diet, chlortetracycline appeared to have a beneficial effect on Campylobacter shedding in the growing and grow-finishing phases of production. However, in the finishing phase when bacitracin replaced chlortetracycline as the dietary antimicrobial, Campylobacter spp. shedding significantly increased in the A+ treatment group and was also two-fold higher compared with the pigs fed antimicrobial-free diets prior to slaughter. Similarly, we observed a significant increase in carcass swabs testing positive for Campylobacter spp. collected in this study near slaughter.

Escherichia coli O157:H7 is a shiga-toxigenic E. coli (STEC) commonly associated with cattle, but rarely associated with swine (Wells et al. 2005; Feder et al. 2007; Farzan et al. 2009). Non-O157 STEC have become a growing concern for human disease, in particular, STEC O26, O103, O111, O121 and O145, which are associated with humans exhibiting haemolytic uraemic syndrome (Friedrich et al. 2002; Mathusa et al. 2010), and immunomagnetic beads have become commercially available for specific strain isolation. Utilizing the USMARC chromogenic medium, we were able to isolate O26, O103, O121 and O145 from our faecal swab enrichments, and isolates with stx, eae and/or hlyA virulence genes were found in 12·2% of the samples. Of the potentially pathogenic O26, O103 and O145 strains, 59·6% of the isolates tested were positive for stx1 or stx2, which indicates that 7·3% of the samples were STEC O26, O103 or O145 positive. In contrast to a previous study in swine where STEC O121 represented 4·6% of the STEC identified from swine faeces (Fratamico et al. 2008), E. coli O121 was isolated on occasion, and none of the isolates were positive for any of the typical pathogenic genes for STEC or EHEC virulence factors. Escherichia coli O26 was the most isolated strain and was highest in the 4-week-old swine fed the antimicrobial-free diet, but most of these strains were negative for stx. Previous work by Cornick (2010) observed that subtherapeutic levels of chlortetracycline in the diet reduced the duration of STEC E. coli O157:H7 shedding in swine. Overall in our study, potentially pathogenic strains of E. coli O26, O103, O121 and O145 were highest early in production and lowest near harvest, regardless of absence or presence of antimicrobial in the diet.

Other STEC E. coli strains may be present in swine faeces (Fratamico et al. 2008; Hutchinson et al. 2011; Prado Martins et al. 2011). Using PCR assays with enrichments, we found that the presence of the stx gene in swine faecal samples was highest prior to slaughter with 58% of the samples positive for the STEC virulence gene. Interestingly, we had significant effects for both dietary antimicrobial and diet phase in the absence of antimicrobials. In the control diet without antimicrobial, there was a marked increase in the presence of stx in faecal enrichments after the switch to finishing rations, which may coincide with the increase in faecal E. coli counts after the change in diets. Subtherapeutic use of chlortetracycline appeared to suppress the presence of stx in enriched faecal samples, whereas subtherapeutic bacitracin increased the presence of stx. The presence of stx2 accounted for nearly all of the stx positives, and the abundance was not surprising, considering that this virulence gene is typically associated with lambdoid prophage and these prophage are readily lysogenic when induced (Kohler et al. 2000). Carbadox and olaqhindox are two swine antimicrobials shown to induce stx2 bacteriophage lysogeny, but chlortetracycline and bacitracin have not been studied in this regard. Regardless of diet or antimicrobial treatment, the percentage of stx-positive samples in this current study was always greater than the 1·4% reported in swine in Brazil (Prado Martins et al. 2011). Hutchinson et al. (2011) reported that 7·9% of swine E. coli isolates collected from clinical submissions were STEC, a percentage greater than they observed in cattle isolates. It appears that the swine gastrointestinal tract might be an important reservoir for stx and STEC. Considering that percentages of stx-positive samples increased late in production but the specific clinically important non-O157 STEC analysed decreased, there is potential for other STEC to have increased and future studies should consider examining total STEC diversity and abundance in production swine.

Changes in Salmonella spp. prevalence were not observed with feeding dietary antimicrobials, but the prevalence for this pathogen was negligible in the current study. The Salmonella enterica enumeration and isolation techniques applied here have been applied to other studies (Schmidt et al. 2012), and our inability to detect Salmonella spp. is likely a consequence of low level of infection in our research swine herd (Wells et al. 2005, 2010) and not a consequence of methods employed.

Previous reports have attributed antibiotic treatment during swine production as a factor driving transmission of Salm. enterica between animals (van der Wolf et al. 2001; Fosse et al. 2009), although a large study observed a higher likelihood for Salmonella in swine fed antimicrobial-free diets (Gebreyes et al. 2006). Salmonella can rapidly bind and invade mucosa and studies have observed increased shedding after antibiotic treatment (Hallstrom and McCormick 2011). Using in vitro techniques to study Salm. enterica colonization, we observed no effect on mucosa binding by strains of Salm. enterica, but did observe a tendency for less internalization of the pathogen strains when dietary antimicrobials were fed. However, it should also be noted that MDR Salm. enterica Typhimurium tended to internalize to a greater extent when dietary antimicrobials were fed, and this statistical tendency became insignificant when we used crypt depth as a covariant in the statistical model. Further research is needed to clarify whether subtherapeutic antimicrobials may alter Salmonella colonization in swine.

Dietary antimicrobials offer the opportunity to improve animal health and prevent disease. However, concern that dietary antimicrobials may result in antibiotic resistance transmission to humans has yielded much debate about their continuous use in animal production. From this current work, subtherapeutic dietary chlortetracycline appeared to reduce Campylobacter spp., some E. coli bearing virulence genes, and the presence of the stx genes in enriched samples of faeces. In contrast, subtherapeutic dietary bacitracin appeared to significantly increase the prevalence for Campylobacter spp. and presence of the stx gene. While previous research indicated that there were changes in the total E. coli and LAB during swine production with phase feeding (Wells et al. 2005), these changes do not appear to be due to dietary antimicrobials being fed as growth promoters.

The current work is the first to provide an in-depth look at major human disease-causing non-O157 STEC in preharvest swine production systems, and while swine harbour a variety of these STEC, the lowest prevalence levels were observed near harvest regardless of the antimicrobial being fed. It is apparent from this work that some dietary antimicrobials fed at subtherapeutic levels may have food safety benefits in swine and consideration needs to be given to these attributes in future studies. Future studies need to not only compare effects of individual antimicrobials but also consider the consequences of antimicrobial phase feeding and examine the potential interactions or antagonisms between antimicrobials in cross-over experimental designs.


Authors would like to thank the USMARC Swine Unit for their expertise and hard work in managing the animals in the project, Donna Griess for her secretarial assistance and Dee Kucera, Shannon Ostdiek and Bruce Jasch for their technical assistance.