In this study, we assess the relationship between the resistome and microbiota, exploiting the resolution provided by shotgun sequence data. Here, we compare the granularity of two data sets – a) contigs and genomes assembled from shotgun metagenome sequencing data and b) OTUs from 16S rRNA gene sequence data as well as gene copy count using qPCR. We target four critical time points on a swine production farm, i.e. farrowing, weaning, growing and finishing, where batch antibiotic treatment (metaphylaxis) was administered. The analysis was motivated by our previous work [11], which concluded that AMR gene levels on this farm were in a state of equilibrium, or saturated, but the antibiotics remained relatively effective in reducing the frequency of infections and mortality. We therefore sought to unravel the microbiome characteristics that underwrite such a phenomenon using amplicon and shotgun sequence data. In doing so, we generate critical insights needed to design strategies that limit welfare impact of disease whilst controlling the spread of antimicrobial resistance on farms.
Characteristics of the gut resistome
The sow microbiome exhibited limited variation in abundance and diversity of antibiotic resistance genes, therefore most of the variation was attributed to piglets [35]. Partitioning this variance suggests that up to 66% may be due to chlorotetracycline and tylosin usage in these piglets. Resistome variability in piglets has previously been reported [35] and its reduction as piglets mature is in part due to natural microbial succession [1, 35]. In this study, we were unable to fully delineate between changes due to the latter and antibiotic usage. Tylosin usage in piglets is known to accelerate microbial succession [35]. However, unravelling this specific aspect was beyond the scope of this study. Fundamentally, we show the direct relationship between microbiota and its resistome, furthermore, the characteristics of the resistome, mode of action for resistance and taxonomic profiles have led us to hypothesise that the marked changes seen at T3 in both piglets and sows may have different precursors.
Changes due to chlorotetracyline and tylosin use
Antibiotic-driven changes of the pig gut microbiome take up 14 days to manifest and may last for months [36, 37]. Here, the piglets had been administered chlortetracycline and acidified water at T2, but the dramatic changes in the selected amplicon resistome, i.e. abundance were observed 28 days later at T3. This may indicate a delayed response attributable to acidified water. At T3, post chlortetracycline administration the full resistome was characterised by an increase in diversity and abundance of tetracycline resistance encoding genes (P > 0.01), for example, tet44, tetW and tetO. Indeed this is reflected as an upregulation of tetracycline ribosomal protection protein was observed, which works by protecting the bacterial ribosome from binding the antibiotic tetracycline [38]. All of this points towards tetracycline-driven changes as reported elsewhere [39].
Beyond this, there was a down regulation of the MDR regulator mechanism which modulates resistance to aminoglycosides and aminocumarins, suggesting a counter selection. This would be expected to impede the functionality of efflux pump mechanisms which confer multi-drug resistance via the bacterial two-component signal transduction system CpxAR [40, 41]. Such an effect would likely compromise the survival of certain microbial populations [42]. Finally, the resistome of piglets at T3 was also characterised by an increase in diversity of genes that encode resistance to glycopeptides and metronidazole, although this indicates co-selection, the mechanism behind this remains unclear. We had expected that the use of antibiotics would result in a reduction of both microbial and resistome diversity, however the use of chlortetracycline was associated with a significant increase in Shannon index [39], suggesting an increase in unique taxa and their abundance. Indeed, this increase is exhibited by a significant reduction of Verrumicrobiaceae and an increase of Bacteroidaceae. Low abundance of the former in piglets under similar conditions has been reported elsewhere [43].
Earlier studies have shown that sub therapeutic administration of tylosin in pigs resulted in a much larger effect than chlortetracycline [35, 43, 44]. In this study, therapeutic levels were used (265.8 mg/PCU tylosin, 103.2 mg/PCU chlortetracycline), and most of the variance in piglets was at T4 after tylosin administration. This could represent a combined effect of both antibiotics, however, if we assume that most of variation attributable to tetracycline was manifested at T3, then tylosin would account for a much larger effect [35, 44]. In this case, counter-selection i.e. chlortetracycline causing a suppression of efflux pump mechanism, may have resulted in microbial population that are more susceptible to antibiotics, leading to a much larger effect when tylosin was subsequently used. The response to this was an upregulation of efflux pumps specific for macrolides and this supports export of chemicals associated with macrolide activity out of the bacterial cell [45]. Interestingly, ermB and lnuC remain abundantly stable across the three time points, which is possibly an indication of legacy use of this antibiotic on this farm. It is also possible that the full extent of tylosin’s impact on this system was in the weeks beyond our selected time points.
The changes to microbiome structure and taxonomic distribution at T4 are comparable with that observed for the resistome, for example; although lower than T3 the Shannon index at T4 is significantly higher than in sows. It is likely that changes at T4 were more taxa-specific, which is why they account for significantly more variation in the resistome [35, 44]. To this effect, a reduction in the abundance of phyla Bacteriodetes, Actinobacteria and Proteobacteria but an increase in Firmicutes might explain the level of change.
Changes due to a potentially undetected infection in sows
The sows did not receive antibiotics during this experiment, however they exhibited marked microbiome structural, resistome and taxonomic changes indicating an undetected incursion. Certain characteristics point towards an infection. Firstly, despite comparable microbial diversity, resistome and taxonomic characteristics were markedly different between sows and piglets at T3 i.e. sows exhibited an up regulation in Class A & B beta-lactamases which catalyse the hydrolysis of the beta lactam ring using serine and zinc-based enzymes respectively [46]. On the other hand, multi-drug efflux pump which are usually down-regulated by upstream pump operon transcription [1, 47]. Such a change is likely to represent an increase in microbial multidrug resistance phenotype for beta-lactams among the sows but the opposite would be expected in piglets [48]. Furthermore, an over representation of genes that encode resistance specifically those that modulate fluoroquinolone resistance, is likely an indication of legacy of use on the farm. Finally, a gradual increase of genes in piglets suggests continuum of exposure via microbial succession hastened or otherwise by antibiotic use [35].
Secondly, viral sequences reveal a marked increase in the abundance of Siphoviridae, Phycodnaviridae, Podoviridae and Smascoviridae. Siphoviridae is associated with Bordetella bronchiseptica, the causative agent of atrophic rhinitis [49]. Indeed, we observe a marked increase (T3) of sequences of the family to which Bordetella species belong. (Fig S7). On the other hand, Phycodnaviridae primarily affects algae [50], suggesting a potential environmental incursion. Smascoviridae, specifically of genus Porprismacovirus has previously been reported in pigs, however its potential to cause disease remains unknown. It is noteworthy that Adenoviridae sequences were also detectable in sows at T3 and these are known to cause mild gastroenteritis [51]. Crucially, the significantly higher abundance of Verrucomicrobiaceae in sows supports the notion of an infection as members in this family i.e. Akkermansia are associated with attenuation of gut inflammatory response [43].
Characterising AMR gene carrying taxa
While it is widely accepted that the vast majority of AMR genes are carried by the uncultured fraction of a microbiome, here we have used 682 near complete genomes across major culturable taxa to show that majority of the AMR genes are carried by Firmicutes and Proteobacteria. This observation supports the notion of a direct relationship between microbiota and its resistome given that these phyla account for 58% of microbial population in this study. Furthermore, we show that a group of Proteobacteria i.e. Escherichia coli and Clostridium sp, were exclusively recovered from piglets and carried a variety of genes encoding resistance to classes of antibiotics i.e. the phenotype of MDR would be expected in this group of bacteria. This further underscores the impact of antibiotic use in this group.
Relevance to swine production
Unravelling gut microbiome characteristics is an essential step in developing strategies that maximise antibiotic utility, nutrient extraction [52], mucosal immune-modulation [52] to improve welfare outcomes [37] for sustainable livestock production systems. To do so, we ought to use methodology that accurately capture these characteristics. Here, we show some differences in the utility of 16S rRNA gene metabarcoding and shotgun metagenomic sequencing and conclude that; a) the former could cost effectively be used to screening for patterns in longitudinal studies, b) and the latter for granular examination of what they represent.
While most studies have focused on bacterial shotgun sequences, here we show that viral sequences can be used to support hypothesis investigations. To this effect we used viral sequences to investigate a suspected infection among sows, and the findings suggest that changes due to infections or otherwise could induce far larger microbiome changes than antibiotic use. It is therefore essential to account for such factors when analysing microbiome data.
Our original hypothesis here was that antibiotic use would restrict microbial diversity due to its bactericidal effect and this would result in an overrepresentation of genes encoding the corresponding resistance. The findings show that antibiotic use in pigs is associated with changes in the microbial structure, abundance, and typically an increase in diversity. The changes in the resistome can take between 14–28 days to manifest and depending on the method of detection, remain detectable for a long time. All this highlights the integral relationship between microbial populations and the genes they carry, and the opportunities and challenges of exploiting them.