The effects of antibiotic use on the dynamics of the microbiome and resistome in pigs

Abstract

Antibiotics are widely used in pig farming across the world which has led to concerns about the potential impact on human health through the selection of antibiotic resistant pathogenic bacteria. This worry has resulted in the development of a production scheme known as pigs Raised Without Antibiotics (RWA), in which pigs are produced in commercial farms, but are ear-tagged as RWA until slaughter unless they receive treatment, thus allowing the farmer to sell the pigs either as premium priced RWA or as conventional meat. Development of antibiotic resistance in pig farming has been studied in national surveys of antibiotic usage and resistance, as well as in experimental studies of groups of pigs, but not in individual pigs followed longitudinally in a commercial pig farm. In this study, a cohort of RWA designated pigs were sampled at 10 time points from birth until slaughter along with pen-mates treated with antibiotics at the same farm. From these samples, the microbiome, determined using 16S sequencing, and the resistome, as determined using qPCR for 82 resistance genes, was investigated, allowing us to examine the difference between RWA pigs and antibiotic treated pigs. We furthermore included 176 additional pigs from six different RWA farms which were sampled at the slaughterhouse as an endpoint to substantiate the cohort as well as for evaluation of intra-farm variability. The results showed a clear effect of age in both the microbiome and resistome composition from early life up until slaughter. As a function of antibiotic treatment, however, we observed a small but significant divergence between treated and untreated animals in their microbiome composition immediately following treatment, which disappeared before 8 weeks of age. The effect on the resistome was evident and an effect of treatment could still be detected at week 8. In animals sampled at the slaughterhouse, we observed no difference in the microbiome or the resistome as a result of treatment status but did see a strong effect of farm origin. Network analysis of co-occurrence of microbiome and resistome data suggested that some resistance genes may be transferred through mobile genetic elements, so we used Hi-C metagenomics on a subset of samples to investigate this. We conclude that antibiotic treatment has a differential effect on the microbiome vs. the resistome and that although resistance gene load is increased by antibiotic treatment load, this effect disappears before slaughter. More studies are needed to elucidate the optimal way to rear pigs without antibiotics.

Keywords: Antibiotic resistance; Metagenomes; Microbiome; Pig farming.

Fig. 1

Sampling scheme of the study. From a cohort of 513 pigs, 103 initial RWA pigs were followed in detail along with 99 progressively enrolled pigs from an initial pool of 410 untreated buffer pigs. One hundred and three pigs were initially followed regardless of treatment (study pigs). For example, in week 2, 84 animals remained RWA, while 14 animals had received antibiotic treatment. In the buffer group, 42 animals had received antibiotic treatment by week 2 and where hence enrolled for sampling downstream, resulting in 140 total pigs registered in that week. Of these 140 pigs, 93 were successfully used for analyses, i.e. had full metadata and complete 16S and qPCR data. Due to logistics and internal movement of the pigs between farm units, some sampling weeks where either under-sampled (week 3) or disturbed by ongoing transfer of pigs between units (weeks 12, 24 and 26) and were hence excluded from statistical testing

Fig. 2

The distribution of key taxa observed in the study, stratified on major phyla and families. The microbiome of each sample was evaluated by sequencing of the V3V4 region of the 16S rRNA gene

Fig. 3

The effect of antibiotics on the porcine gut microbiome. The microbiome of each sample was evaluated by sequencing of the V3V4 region of the rRNA gene. a All samples of the study coloured by sampling time. Weeks marked with * in the legend are not included in the other panels due to under-sampling. b–h Animals were stratified at each time point into untreated (black) and antibiotic treated groups (red). Significance was evaluated by PERMANOVA at each time point adjusting for pen and variance described by treatment status is represented by the R²-value and corresponding p-value. Ellipses represents the 95% standard error of the centroids of each group. All panes are represented by the first and second nMDS-axis. Since the axes of nMDS are arbitrary, they are removed for brevity

Fig. 4

The effect of antibiotics on the porcine gut resistome. The resistome of each sample was evaluated by high-throughput qPCR on 82 ARGs. a All samples of the study coloured by sampling time. Weeks marked with * in the legend are not included in the other panels due to undersampling. b–h Animals were stratified at each time point into untreated (black) and antibiotic treated groups (red). Significance was evaluated by PERMANOVA at each time point adjusting for pen and variance described by treatment status is represented by the R²-value and corresponding p-value. Ellipses represents the 95% standard error of the centroids of each group. All panes are represented by the first and second nMDS-axis. Since the axes of nMDS are arbitrary, they are removed for brevity

Fig. 5

The effect of antibiotic treatment on a set of 176 pigs from 6 different farms taken at termination (26 weeks) at a slaughterhouse. A The microbiome of each sample as evaluated by sequencing of the V3V4 region of the rRNA gene. B The resistome of each sample as evaluated by high-throughput qPCR on 82 ARGs. Significance was evaluated by PERMANOVA and showed a significant effect of farm, but not of treatment status

Fig. 6

Resistance gene abundance across the study. The overall abundance of ARGs presented as the median sum of resistance genes in each pig per sampling time. a Samples are considered treated or untreated and b samples are considered on a time-of-treatment basis. Significant difference between treatment group and control at sampling time by Kruskal–Wallis test followed by Conover’s test is denoted by * at each time point. T01-T05 denotes treatment at the corresponding week. UT: untreated

Fig. 7

Co-occurrence networks of bacterial and ARG abundance for untreated and treated animals at week 2, 4, 5 and 6, highlighting how genes and bacteria associate differently depending on treatment. Networks were generated using centered log-transform normalization and a correlation cut-off of 0.7. The IncN rep/IS26 cluster of differentially abundant genes is highlighted by red circles. Edges are scaled according to correlation