The impacts of viral infection and subsequent antimicrobials on the microbiome-resistome of growing pigs

Abstract

Background: Antimicrobials are used in food-producing animals for purposes of preventing, controlling, and/or treating infections. In swine, a major driver of antimicrobial use is porcine reproductive and respiratory syndrome (PRRS), which is caused by a virus that predisposes infected animals to secondary bacterial infections. Numerous antimicrobial protocols are used to treat PRRS, but we have little insight into how these treatment schemes impact antimicrobial resistance (AMR) dynamics within the fecal microbiome of commercial swine. The aim of this study was to determine whether different PRRS-relevant antimicrobial treatment protocols were associated with differences in the fecal microbiome and resistome of growing pigs. To accomplish this, we used a metagenomics approach to characterize and compare the longitudinal wean-to-market resistome and microbiome of pigs challenged with PRRS virus and then exposed to different antimicrobial treatments, and a group of control pigs not challenged with PRRS virus and having minimal antimicrobial exposure. Genomic DNA was extracted from pen-level composite fecal samples from each treatment group and subjected to metagenomic sequencing and microbiome-resistome bioinformatic and statistical analysis. Microbiome-resistome profiles were compared over time and between treatment groups.

Results: Fecal microbiome and resistome compositions both changed significantly over time, with a dramatic and stereotypic shift between weaning and 9 days post-weaning (dpw). Antimicrobial resistance gene (ARG) richness and diversity were significantly higher at earlier time points, while microbiome richness and diversity were significantly lower. The post-weaning shift was characterized by transition from a Bacteroides-dominated enterotype to Lactobacillus- and Streptococcus-dominated enterotypes. Both the microbiome and resistome stabilized by 44 dpw, at which point the trajectory of microbiome-resistome maturation began to diverge slightly between the treatment groups, potentially due to physical clustering of the pigs. Challenge with PRRS virus seemed to correspond to the re-appearance of many very rare and low-abundance ARGs within the feces of challenged pigs. Despite very different antimicrobial exposures after challenge with PRRS virus, resistome composition remained largely similar between the treatment groups. Differences in ARG abundance between the groups were mostly driven by temporal changes in abundance that occurred prior to antimicrobial exposures, with the exception of ermG, which increased in the feces of treated pigs, and was significantly more abundant in the feces of these pigs compared to the pigs that did not receive post-PRRS antimicrobials.

Conclusions: The fecal microbiome-resistome of growing pigs exhibited a stereotypic trajectory driven largely by weaning and physiologic aging of the pigs. Events such as viral illness, antimicrobial exposures, and physical grouping of the pigs exerted significant yet relatively minor influence over this trajectory. Therefore, the AMR profile of market-age pigs is the culmination of the life history of the individual pigs and the populations to which they belong. Disease status alone may be a significant driver of AMR in market-age pigs, and understanding the interaction between disease processes and antimicrobial exposures on the swine microbiome-resistome is crucial to developing effective, robust, and reproducible interventions to control AMR. Video Abstract.

Keywords: Antimicrobial resistance; Metagenomics; Microbiome; Porcine reproductive; Respiratory syndrome; Swine.

Fig. 1

100% stacked graphs depicting a relative abundance of antimicrobial resistance (AMR) classes (A) and microbial genera (B), grouped by time point and treatment group (Min = Minimal; Mod = Moderate; Int = Intensive). AMR classes with <0.1% relative abundance and taxa phyla with <5% relative abundance across all samples were grouped as “Others”. Relative abundance was based on alignment counts normalized using cumulative sum scaling (CSS)

Fig. 2

Sampling time point is associated with significant differences in fecal resistome and microbiome composition. A Dendrogram showing linkage clustering of Bray-Curtis (BC) dis(similarities) of fecal resistome composition at the ARG level, colored by time point and treatment group. Non-metric multidimensional scaling (NMDS) ordination based on BC (dis)similarities (stress=0.076) by B time point (ANOSIM—analysis of similarities P=0.001, PERMANOVA—permutational multivariate analysis of variance R²=53.9%, P=0.001) and C treatment group (ANOSIM P=0.07, PERMANOVA R²<1%, P=0.22). NMDS ordination based on BC (dis)similarities (stress=0.10) of microbial genera by D time point (ANOSIM P=0.001, PERMANOVA R²=42.6%, P=0.001) and E treatment group (ANOSIM, P=0.32, PERMANOVA R²<1%, P=0.11). Ellipses indicate the 95% confidence interval for distance from the centroids of each respective group of samples

Fig. 3

Resistome and microbiome alpha-diversity. Genus-level Shannon’s diversity (A), Pielou’s evenness (B), and richness (C). ARG group-level Shannon’s diversity (D), Pielou’s evenness (E), and richness (F). Horizontal black lines indicate pairwise comparisons between timepoints with statistically significant differences at the P < 0.01 (*), P < 0.001 (**), and P < 0.0001 (***), based on generalized linear mixed modeling

Fig. 4

Enterotype dynamics over time and between treatment groups. Dirichlet multinomial mixture (DMM) models of genus-level microbiome alignments were used to assign samples into one of eight clusters based on the lowest Laplace approximation. A Transitional model showing assignment of fecal samples to DMM cluster (y-axis) and stratified by time point (x-axis). The size of circles is proportional to the number of samples contained in each DMM cluster, and edges are weighted by transition frequency. B Stack bar of cluster stratified by the treatment group (colored)

Fig. 5

Log-fold change (logFC, x-axes) of ARG features A each treatment group (minimal, moderate, and intensive) comparing sequential sampling time points (i.e., positive logFC values indicate higher abundance in the later time point compared to the earlier time point) and B between treatment groups (minimal, moderate, and intensive) for each time point. Each dot represents the ARG feature in the group, and red dots represent ARG that are significantly different between comparison groups (i.e., logFC≥±1, BH adjusted P< 0.05)

Fig. 6

Dirichlet multinomial mixtures (DMM) samples into one of three clusters from the entire resistome composition at ARG level based on the lowest Laplace approximation. A Heat map showing the abundance (square root transformed) of the 30 most dominant ARG per DMM cluster, B transitional model showing the progressive change of sample through DMM cluster/resistotypes per each sampling point from wean-to-market across all treatments, and C clusters stratified by treatment group (colored). Time points are on the x-axis, and resistotype is represented on the y-axis. The size of the circle is proportional to the number of samples contained in each DMM cluster, and nodes are colored and edges are colored by transition frequency

Fig. 7

Binary heatmap (red=present, gray=absent) of resistance mechanisms for each resistance class by treatment group for each sampling time point (list of each mechanism by AMR class is presented in Additional file 1: Table S1)