Resistance to change: AMR gene dynamics on a commercial pig farm with high antimicrobial usage

Group antimicrobial administration is used to control disease in livestock, but we have little insight into how this impacts antimicrobial resistance (AMR) gene dynamics. Here, a longitudinal study was carried out during a single production cycle on a commercial pig unit with high historic and current antimicrobial usage. Quantitative PCR, 16S rRNA gene metabarcoding and shotgun metagenomic sequencing were used to track faecal AMR gene abundance and diversity and microbiome alpha diversity. Shotgun metagenomic sequencing identified 144 AMR genes in total, with higher AMR gene diversity present in young pigs compared to dry sows. Irrespective of in-feed antibiotic treatment or changes in microbiome diversity, mean AMR gene copy number was consistently high, with some AMR genes present at copy numbers comparable to the bacterial 16S rRNA gene. In conclusion, AMR gene prevalence and abundance were not influenced by antibiotic use, either during the production cycle or following whole-herd medication. The high levels of certain genes indicate they are widely disseminated throughout the microbial population, potentially aiding stability. Despite the high and relatively stable levels of resistance genes against the main antimicrobials used, these compounds continue to control production limiting diseases on this unit.


Current antimicrobial administration
During the study period, the following routine group medication regimens were used: toltrazuril (30 mg/head oral) at 4 days old to control Isospora suis, zinc (2500 ppm in feed) between 4 and 6 weeks old to control post-weaning colibacillosis, acidified water (Baynes Evacide S 0.2%) to control postweaning colibacillosis between 3 and 7 weeks old, chlortetracycline (300 ppm in feed) from 6 to 8 1/2 weeks old to control Mycoplasma hyopneumoniae and Mycoplasma hyorhinis and tylosin (100 ppm in feed) from 8 1/2 weeks old until slaughter to control Mycoplasma hyopneumoniae, Mycoplasma hyorhinis, Actinobacillus pleuopneumoniae and Lawsonia intracellularis.
During the partial depopulation, the dry sows received 1500 ppm chlortetracycline and 500 ppm tiamulin in-feed, whilst the nursing sows received 1875 ppm chlortetracycline and 625 ppm tiamulin.
In the three months that included this partial depopulation, total antimicrobial use increased to 582.8 mg/PCU, which then declined to 32.3 mg/PCU in the three months after the partial depopulation.

Faecal sampling
On W1, faecal drop samples were collected from the floor of six farrowing crates containing pregnant sows (n = 6). Between W2 and W4, both sow and piglet faecal drop samples were obtained from the same farrowing crates as W1. On W5, all piglets were weaned and mixed into three groups prior to movement into the weaning house. From W5 to W13, pooled faecal drop samples were taken from each of these three pens to capture pen-level dynamics. On W14, these pigs were moved into the grower/finisher house and remained in the same pen formation as in the weaner house. Thereafter, from W14 to W25, pooled faecal drop samples were taken weekly from each of these pens until slaughter.

Quantitative PCR selection
The selection of AMR genes for targeted quantification (n = 5) was also based on the results of an initial end-point PCR screening of a sub-sample of DNA extracts obtained from the final sampling point (faecal samples = 6). A panel of 30 genes were selected, on the basis that these genes were of biological relevance to the historic use of antimicrobials on the pig unit and of importance to both veterinary and human medicine (see Supplementary Materials 2 for list of target genes and primer sequences). Genes which were amplified from more than 50% of the faecal samples were shortlisted for qPCR analysis.

Quantitative PCR methodology
qPCR mixtures were set up using Brilliant III Ultra-Fast qPCR Mastermix (Agilent Technologies, United States), reference dye (Agilent Technologies, United States) and the primers and probes listed in Supplementary Material 2. Each reaction was carried out in triplicate in a final volume of 20 l, containing 1 l of extracted DNA. Twenty-four samples were run per 96-well plate, which also included DNA standards (at concentrations ranging from 10 7 to 10 1 gene copies per l) and a no-template control (nuclease-free water). Absolute quantification was carried out using a Stratagene MX3005P qPCR System (Agilent Technologies, UK) using the following fast, two-step cycling conditions: 95°C (5 minutes), followed by 40 cycles of amplification at 95°C (15 seconds) then 60°C (30 seconds).
Standard curves were constructed from the threshold cycle (CT) values using the Stratagene MxPro Software (Agilent Technologies, UK) and within this software, the calculated gene copy number per l for each of the samples was generated, treating each of the three replicates individually. Any samples which fell beneath the limit of detection (i.e. 10 1 copies per µl DNA, equivalent to 3.3 -4.9 log10 copies/g DM) were re-run to confirm the findings. These values were then exported in Microsoft Excel spreadsheet format and the arithmetic means for the technical replicates calculated. These values were then converted into gene copy number per gram of dry matter (using the % DM values calculated) and log10-transformed for data visualisation and statistical analysis.

16S rRNA gene metabarcoding
Six library pools were compiled using equimolar concentrations of DNA from each sample. A mock bacterial community (20 Strain Even Mix Genomic Material ATCC®MSA-1002, ATCC, United States) and a reagent-only control (generated by passage of nuclease-free water throughout the library preparation process) were included in each pool to assess background contamination and sequencing error rate. Using the mock bacterial community sequences, the mean sequencing error rate was calculated as 0.01%.
The pools were submitted to the sequencing centre (Edinburgh Genomics, United Kingdom) where the pools were quantified using the Quant-iT™ PicoGreen® double-stranded DNA Assay Kit (Thermo Fisher Scientific, UK) to ensure sufficient yield for sequencing. Sequencing was carried out using the Illumina MiSeq platform (Illumina, United States), using V2 chemistry and producing 250 bp pairedend reads.

Metagenomic sequencing
Illumina TruSeq DNA Nano libraries were prepared using the submitted faecal DNA extracts and sequencing was carried out using the HiSeq 4000 platform generating 150 bp paired-end reads (Illumina, United States).
Host DNA was removed from the raw reads by mapping to the Sus Scrofa reference genome GCA_000003025 version 11.1 and Phix DNA (PhiX 174) was removed using the run_contaminant_filter.pl script which is part of the Microbiome Helper suite. Read quality control was carried out using trimmomatic. These paired end reads were then mapped to MEGARes, a hand-curated AMR gene database, using the paired-end option of BWA. The resistome profiling was carried out using the ResistomeAnalyzer function in MEGARes with the gene fraction threshold set at 90. In order to obtain a read per kilobase (RPK) normalised abundance count, the Humann2 script which is part of the Humann2 pipeline was used to analyse the SAM file produced from the mapping step above. The resultant gene family output was then normalised using the humann2_renom_table script. The generated tsv and csv files were then used to analyse AMR gene abundances in R version 3.5. The downstream analysis of AMR gene diversity and data visualisation were carried out using the ggplot2 package in R. * Designed as part of this study. Primers/probes were designed using Primer BLAST (NCBI), with the latter being checked for specificity against the GenBank nr database.

Supplementary Materials 3 -Statistical outputs
Young pig accommodation: Assessment of temporal variation of mean gene copy number and mean diversity indices by analysis of variance, presented with the mean standard error of difference (SED) for each model. The models excluded samples from W1 (pens only contained pregnant sows at this time point) and samples from W2-W4 when the young pigs were still grouped by litter and were not yet assigned to the rearing pens. Week