Resistomes from oxytetracycline-treated pigs are readily transferred to untreated pen mates

Abstract

Pork is currently a major part of Danish food export and is also a key dietary source of protein across the world. Industrial pork production, however, comes with high antibiotic usage in many countries, including Denmark. This has created consumer demand for meat Raised Without Antibiotics (RWA). Previous work has demonstrated that levels of antibiotic resistance genes (ARGs) are indeed increased in antibiotically treated animals, but also suggest that these ARGs are transferred to untreated pen-mates. In a Danish commercial farm, we studied four groups of physically separated pigs: one group of only antibiotic treated pigs (n = 20), one group of only untreated pigs (n = 30 total, n = 15 analysed), and one group combining treated (n = 15) and untreated pigs (n = 15). These groups were followed for 16 weeks during which all pigs were profiled for both their faecal microbiome (through 16 S rRNA gene sequencing) and resistome (by use of a high-throughput qPCR platform targeting 82 ARGs and their variants). We found that the resistome of treated pigs was substantially enriched in resistance genes compared to untreated pigs but, importantly, observed that untreated pigs co-reared with treated pigs had levels of resistance genes approaching their treated pen mates, suggesting that the treated enterotype is readily transferred to the untreated animal. From this, we conclude that mixing of treated and untreated pigs causes spill-over of antibiotic resistant bacteria and/or resistance genes from treated pigs when these are co-reared. To optimize RWA production, treated and untreated pigs should be physically separated to limit the proliferation of ARGs.

Introduction

Although antibiotics are essential in the healthcare of both humans and animals, their usage are bound to result in microbes that are resistant to them [1]. In recent years, the use of antibiotics in animal production has come under increasing scrutiny, namely because antibiotic resistance in livestock is increasing in relevance [2]. This, in turn, inevitably increases the risk of transferring antibiotic resistant bacteria and/or resistance genes from livestock to humans. Indeed, zoonotic transmission of resistant bacteria has been demonstrated for several pathogens, such as E. coli, Salmonella, Campylobacter, and methicillin-resistant Staphylococcus aureus, although the general consensus is that the contribution of food producing animals in the spread of antibiotic resistance in human pathogens is still very low [3]. In any case, the health of animals and humans are fundamentally intertwined and should be ideally be considered with a one-health perspective in mind [4]: decreasing antibiotic usage in animals will be beneficial for humans and vice versa.

An initiative in this regard is to reduce antibiotic usage in industrial husbandry, i.e. the concept of ‘Raised Without Antibiotics’ (RWA). Although there are different interpretations of this concept, the general consensus is “no antibiotics ever” meaning that RWA labelled meat originates from an animal having never been treated with antibiotics [5]. Indeed, results from beef production have indicated lower levels of antibiotic resistance at slaughter in the intestinal microbiome of RWA animals compared to conventional animals [6, 7]. Some concerns, however, have been raised regarding animal welfare in RWA systems due to the potential reluctance to treat sick animals [7], although the scale and/or existence of this issue remains to be seen.

In the Danish concept for RWA pig production, all pigs in a RWA herd are ear tagged at birth [8]. A pig continuing to slaughter with an intact earmark will then net a premium price, whilst a non-earmarked pig – i.e. a treated one – is considered merely conventional. This traceability of antibiotic treatments in the RWA concept poses a unique opportunity to track antibiotic treatments in commercial production systems and to examine the transmission of antibiotic resistant bacteria between treated and untreated pigs in a fully operating pig farm.

Previous studies investigating the RWA concept in two Danish farms showed that 32–36% of the RWA-pigs were treated with antibiotics due to illness at 0–12 weeks of age [9], in contrast to practically 100% in conventional farms. As part of the production in conventional farms, pigs are mixed in pens after weaning and pigs treated with antibiotics are most often reared in the same pen as untreated pigs.

An important scientific question is how RWA production influences the pig microbiome, i.e., the bacterial communities inhabiting the pig. Of special interest is the intestinal microbiome, which facilitate digestion [10] and prevents pathogens from causing disease [11]. The pig microbiome is affected by multiple conditions such as breed, host genes, age, gender and castration of male pigs [12]. Across the lifetime of a pig, both the composition and diversity of the microbiome will change [13], with the most marked alteration taking place during weaning. Here, the diet of the pig is changed from sow’s milk to solid feed, causing a relative decrease of bacteria from the Lactobacillaceae family [14], which instead is replaced by genera like Catenibacterium, Prevotella, and Yersinia owing to the inherent ability of these genera in digesting the more complex carbohydrates found in the plants they are fed [15].

Apart from age and weaning, treatment with antibiotics has been shown to lead to both short and long-term effects on the composition and diversity of the microbiome [16]. Early-life treatment in particular has been reported to have a lasting compositional change of the pig microbiome, namely by reduction of diversity [17]. This can be problematic, as a highly diverse microbiome contributes to resilience towards pathogens and therefore the health of the pig [11].

As an inherent characteristic of a microbiome, the resistome can be defined as the set of antibiotic resistance genes (ARGs) residing within this microbiome which, hence, confer the collective antibiotic resistance phenotype. As would be assumed, the resistome is altered significantly both in abundance and diversity when weaning pigs were administered oxytetracycline [18], an observation mirrored in large-scale cross-country analyses, in which ARGs of pigs and chickens appear more abundant in countries with a high antibiotic usage [19]. The methodology for this analysis is however of key importance – counts of indicator bacteria, metagenomics and qPCR have all been used, making direct comparisons difficult. qPCR is the only method to directly quantify each ARG, but this method does not have the breadth of metagenomics unless one can analyse multiple genes in parallel. This approach was developed in a previous study, in which a high throughput qPCR Fluidigm platform was used to target 82 resistance genes simultaneously [16]. This method builds on prior investigations, namely the work of Græsbøll et al., in which a clear effect of antibiotic treatment regimens was observed on resistance gene composition [20].

The current study builds upon prior work by Tams et al. 2023 [16], in which a cohort of RWA pigs – both treated and not - in an industrial farm were followed across their lifetime. When analysing both the microbiome and resistome of these pigs, we found a robust and clear effect of antibiotic treatment, although this effect was small in size and transient in nature. Since the pigs are co-reared and coprophagic, we hypothesized that the treated entero-/pheno-type is transferred to the untreated animals, hence obstructing the microbiological aims of the RWA concept.

To investigate this in more detail, we designed an experimental study where treated and untreated animals were mixed in a controlled herd setting and observed to which degree they exchanged their microbiome and resistome. When comparing these groups with two physically isolated groups of treated and non-treated animals, we observed the co-reared but untreated animals taking on the enterotype of their treated pen-mates rather than the untreated controls.

Methods

Farm characteristics

The study was carried out in a Danish RWA herd with 670 sows, 4,700 nursery pigs and 3,200 finisher pigs on site. All pigs from a weekly batch were sampled 4 days before the study start (week 0) as they entered the weaning unit in order to determine the baseline microbiome of the subset of pigs which would eventually be divided into the study groups. In the farrowing unit, piglets treated with antibiotics at any time point were moved to a separate, isolated farrowing room. Sows treated with antibiotics were likewise moved and isolated. The untreated piglets included in the trial had thus never been in physical contact with a sow or pen mates that had received antibiotic treatment in the farrowing unit. No washing nor disinfection was performed between farrowing batches, which is the standard procedure in this particular herd. At weaning, the pigs were moved into a weaning unit, which had been cleaned by high pressure washing and disinfected with lime followed by an empty drying out period of 2 days.

Experimental design

Four groups of pigs were included for microbiome and resistome analyses: group Treated (n = 20), in which all animals were treated, group Untreated (n = 15), in which no animals where treated, and lastly animals from a mixed pen, containing two physically co-reared subgroups consisting of groups Mix-treated (n = 15), in which all animals were treated, and group Mix-untreated (n = 15), in which no pigs where treated. This design allowed us to compare not only Treated with Untreated, but also to elucidate any transfer of a treated phenotype from Mix-treated to Mix-untreated. All antibiotic treatment during the study was done with oxytetracycline (Engemycin^®Vet, 100 mg/ml intramuscularly for three consecutive days). The pigs included in the study were groupwise located in one part of a double pen with a shared trough (Fig. 1). The pigs located in the other part of the double pen were chosen to have a similar status regarding antibiotic treatment as the study pigs in order to reduce the potential confounding impact of nasal contact on the microbiome and resistome.

Fig. 1

Illustration of the pig herd weaning room hosting the 3 trial pens. Animals in the Treated pen were treated with antibiotics, animals in the Untreated pen were not, and animals in the mixed pen belonged to one of either subgroups of treated (Mix-treated) and untreated animals (Mix-untreated). All trial pens had shared throughs and hence contact to neighbouring pens with pigs of similar treatment status. Animals not part of the study were reared as usual in the stable as per standard farm practices

The weaning unit of the farm consisted of eight double pens designed to contain up to 60 pigs (Fig. 1) of which three double pens (Treated, Mixed and Untreated) were reserved as trial pens for this study (Fig. 1). These double pens were characterized by two pens sharing feed and water trough with possibility of nasal contact to non-study animals of similar status (contact animals). Solid walls were temporarily built to avoid physical contact between Treated, Mixed and Untreated pens (Fig. 1). A separate pair of boots were placed outside each trial pen to minimize faecal contamination between pens by trial and herd personnel. The remaining pens in the room housed untreated, non-study pigs of similar age.

Experimental procedure

At weaning, 30 untreated pigs were transferred from the farrowing unit and distributed across pens within the nursery room, leaving out the pens dedicated for Treated and Mixed pigs, which were instead left empty.

During the first two weeks in the nursery room, all sick pigs needing antibiotic treatment (n = 6), were transferred to the Treated pen and treated with oxytetracycline. At week 3 post-weaning, the final study groups were established by distributing 27 smaller pigs (average weight 5–6 kg) with affected body condition from the nursery room to the Treated (n = 12) and Mixed (n = 15, group Mix-treated) pens and treating all 27 pigs with oxytetracycline. In addition, 15 smaller pigs were transferred to the Mixed pen and not treated with oxytetracycline (group Mix-untreated). At the same time, 15 untreated pigs were moved to the pen adjacent to the research animals in the Mixed pen (Untreated-contact). Due to animal death and non-study treatment, the exact number of animals varied across time (Table S1). Regardless of these additions, only pigs initially chosen for analysis were included for final analysis.

At 8 weeks post weaning, the pigs were moved to a cleaned and disinfected finisher unit. In the finisher unit, pigs were kept in the same pen-groups as in the nursey room except for mixing of pigs from both Untreated pens.

Experimental procedures

Sampling

Sampling of the microbiome for molecular analysis was done with faecal swabs using sterile Dryswab^® Fine Tip MW113 swabs (Medical Wire Equipment, Corsham, UK). We sampled before weaning (in the farrowing unit) and in weeks 3, 5, 6 and 7 post-weaning (in the weaning unit) and at 10, 13, and 16 weeks post-weaning (in the finisher unit). Faecal swabs were transferred to 3.6 mL cryotubes containing 1.5 mL PBS and stored on ice until processing. Subsequently, the tubes were vortexed vigorously and the liquid contents were stored at -80 °C for further processing within 24 h after sampling.

DNA extraction

Samples were DNA extracted from 350 µL faecal sample on columns DNeasy PowerSoil Pro Kit (QIAGEN, Hilden, Germany) according to the manufacturer guidelines.

High throughput qPCR

A high-throughput qPCR array was used to quantify 82 individual ARGs as described in Tams et al. 2023 [16]. First, DNA concentrations were measured on a NanoDrop ND-1000 Spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA) and diluted to < 10 ng/µl in nuclease-free water (Qiagen, Hilden, Germany) and were until further processing stored at − 20 °C. Then, qPCR data was obtained on the Fluidigm HD Biomark system (South San Francisco, CA, USA) with EvaGreen as the fluorescent marker. PCR was carried out as follows: thermal mix, at 50 ⁰C for 2 min, 70 ⁰C for 30 min, 25 ⁰C for 10 min; UNG and Hot Start 50 ⁰C for 2 min, 95 ⁰C for 10 min; 35 Cycles of, 95 ⁰C 15 s, 60 ⁰C for one min; followed by a melting phase 60 ⁰C for 30 s lastly the temp was raised to 95 ⁰C by 1 ⁰C per 3 s.

Primers for all ARGs are described in detail in Tams et al. 2023 [16], and represent a variety of tetracycline-, beta-lactam-, sulfonamide-, aminoglycoside-, zinc-, trimethoprim-, lincosamide-, macrolide, and streptogramin resistance genes. Moreover, one plasmid replication initiator protein and three IS element associated with plasmids carrying ARGs where included. Controls are described in detail in Tams et al. [16], and were included on each chip. Positive controls included artificial amplicons as well as relevant strains and faecal samples, whereas negative controls were buffers and empty wells. The abundance of each ARG within each sample was normalized by the 16 S qPCR abundance of the same sample [16].

16SrRNA sequencing metataxonomics

Sequencing of the 16 S rRNA gene was done as described by [21]. Briefly, DADA2 was used for classification as described in Callahan et al. 2016 [22], using the Silva database as a reference and the species-level training set (v. 138). The total number of samples was 551 (controls excluded), the mean input read count was 38,609 (range 14,785 − 106,580), the per-sample mean filtered and trimmed read count was 32,908, the read count after merging in DADA2 was 28,540 and the non-chimeric read count was 26,002.

Statistics

All statistics were done in R v4.1.1 and all multivariate statistics were performed by use of the vegan package [23]. For all multivariate analysis, the metataxonomic data were considered at the ASV level. ARG data were analysed after normalization by use of within-sample 16 S rRNA qPCR as described above.

Pairwise analysis was either conducted by comparison of all treated vs. all untreated animals or by pairwise comparison of all four groups. This analysis was furthermore extended to the 9 time points of the study. Week 0 represents the pigs in the farrowing unit before antibiotic treatment and separation into pens in the weaning unit. Week 3 represents the pigs during treatment (2nd day of a 3-day treatment regimen). The subsequent weeks (4, 5, 6, and 7) were from the weaning unit and the last weeks (10, 13, 16) were collected in the slaughter unit.

Univariate analysis

In order to quantify the overall load of ARGs, the normalized resistance gene abundances were first summed for each pig, after which the median was calculated for each group of pigs. As the qPCR data was highly zero-inflated and heteroscedastic, we used the non-parametric Kruskal-Wallis test. Furthermore, a pairwise Dunn’s test was used to specifically compare the sets Treated vs. Untreated, Mix-treated vs. Mix-untreated, Untreated vs. Mix-untreated and Treated vs. Mix-treated. Individual ARGs were tested in a similar scheme if at least one group had more than 50% non-zero values.

Individual bacteria were tested for difference in relative abundance using ANCOM-BC at the genus level. Further analysis was focused on the contrasts between Mix-treated and Mix-untreated, which was deemed significant if a genus was significantly differentially abundant at more than two timepoints.

Multivariate analysis

Both metataxonomic and ARG data were separately analysed by Permutational Multivariate Analysis Of Variance (PERMANOVA) with 200,000 permutations using the overall model $\:\left(Y\:=\:Time\:+\:Group\:+\:Time:Group\right)$ on all time points, followed by individual within-week models $\:\left(Y\:=\:Group\right)$. Since the within-week PERMANOVA cannot specify which individual groups differ from one another, pairwise PERMANOVAs were performed within-week on groups Treated vs. Untreated (to elucidate the difference between treated and untreated pigs) as well as one on Mix-treated vs. Mix-untreated (to elucidate the difference between treated and untreated pigs in the same pen). When the (adjusted) P-value of a comparison was below 0.05, the R² value (i.e., the multivariate variance explained by Group) was used as the main metric for inferring effect size. The P-values of the 9 individual within-week PERMANOVAS were adjusted by the Benjamini-Hochberg procedure. Visualisation of multivariate data was generated using non-Metric Multidimensional Scaling (nMDS).

All code for analysis is available at https://github.com/mikaells/Transfer_study. All sequencing data is available as bioproject PRJNA1023074.

Results

A total of 349 live born piglets from 19 sows were included in the study. The pigs selected to be followed throughout the study were distributed in the treatment groups Treated, Mix-treated, Mix-untreated and Untreated and constituted 20, 15, 15, and 15 pigs respectively.

The load of ARGs is driven by antibiotic treatment but can be transferred to untreated penmates

First, the pattern of ARG load as a function of group was investigated. We estimated the overall ARG load by summing all resistance genes within each sample and then presenting the median. To investigate at which time-point the treatment groups significantly differed, we used non-parametric statistics (Fig. 2). Treatment status had a substantial effect on resistance gene load. This was most evident by an extensive and significant elevation immediately after treatment in week 3 in the two groups treated with antibiotics, i.e. groups Treated and Mix-treated, as compared to the untreated groups Untreated and Mix-untreated. This effect was apparent until week 10 (Fig. 2a). When investigating the groups in more detail, we observed the control group, Untreated, to have low levels throughout the study, contrasting strongly with the Treated group, which exhibited a marked elevation from treatment start to week 10. The treated animals in the mixed group, Mix-treated, had levels of resistance indistinguishable from group Treated. Of main interest, however, was group Mix-untreated, which was indistinguishable from group Untreated in week 3 and 4 but reached a level of resistance hovering between the treated groups and the control group in weeks 5 to 7. This key result suggests that antibiotic resistance is progressively transferred from treated to untreated animals when sharing a pen. At week 10, however, Mix-untreated was significantly different from group Untreated and instead reached levels of resistance close to groups Mix-treated and Treated. At weeks 13 and 16, no statistical difference was observed between any of the groups, suggesting that the final resistance load in slaughter-ready pigs is essentially independent of the treatment regime they have previously received.

Fig. 2

The sum of ARGs in each sample stratified by either (a) treated vs. untreated pigs or (b) by treatment groups. The vertical line denotes treatment time. For panel A, * denotes statistical difference between treated and untreated pigs. In panel B, ‘a’ denotes difference between Treated & Mix-treated vs. Untreated & Mix-treated, ‘b’ denotes difference between Treated & Mix-treated vs. Mix-untreated vs. Untreated, ‘c’ denotes difference between Treated vs. Mix-treated vs. Mix-untreated vs. Untreated and ‘d’ denotes difference of Treated & Mix-treated & Mix-untreated vs. Untreated

Individual ARGs follow total ARG load

Having established the overall load of ARGs is a function of treatment status as well as being transferable, we next aimed at determining which specific ARGs might be key in this dynamic. We here observed ten out of a total of 82 genes to be significantly affected by treatment (Fig. 3).

Apart from a large carry-over of aminoglycoside resistance from the farrowing unit, which remained as a small but continuous elevation in all treated groups as compared to Untreated animals, seven out of the 10 genes were tetracycline resistance genes, corresponding well to the antibiotics administered in this study. Of special interest was the pattern of the genes tetO and tet [32], all of which appeared particular abundant in the Mix-untreated groups, suggesting that organisms or genetic elements containing these genes are particularly easy to transfer. These results also highlight the relevance of considering the total ARG load instead of focusing on individual genes.

Fig. 3

The abundance of individual resistance genes normalized by the 16 S rRNA gene. Only genes with at least one date significantly different between 2 groups or more are included. *: significant difference between at least two groups by the Kruskal Wallis test

Multivariate analysis of ARGs shows a time-dependent transfer of the resistome

Next, we focused on the multivariate pattern of the ARGs. Since we had observed a gradual convergence of ARG load between untreated pigs and their treated pen-mates, we hypothesized that the resistome composition of Mix-untreated pigs would gradually assume the composition of Mix-treated pigs. To address this, we next investigated the dynamics of ARG composition as a function of group with multivariate statistics. Using an overall model including both treatment-status and time, we observed a substantial effect of Time (R² = 0.18), as is expected in a time-series of growing animals. More importantly, however, were the 5.11% variance explained by the treatment status, along with a significant interaction of treatment and time (R² = 2.02%, p = 1*10^− 5), suggesting that the effect of treatment depends on the age of the animal. This leads us to focus the analysis on individual weeks by use of nMDS and PERMANOVA, and here we observed a much more evident effect of treatment (Fig. 4 & Table S2). Here, the groups were indistinguishable before treatment (Week 0), followed by a clear and significant clustering of each treatment group from weeks 3 to 13, the most remarkable being a strong separation of the Untreated group from all others. Of especial interest was the alignment of the two Mixed groups with their respective control groups. Specifically, group Mix-untreated clustered with group Untreated and group Mix-treated with group Treated, which is particularly evident in week 3 (Fig. 5b). From week 4 to 16, the Untreated and Treated groups remain significantly different from one another, corresponding to their difference in treatment status. Remarkably, groups Mix-untreated and Mix-treated remain different from one another from weeks 3 through 7, although group Mix-untreated progressively moves towards the now indistinguishable duo of treated groups. By week 10, the Mix-untreated group, despite having never been treated with antibiotics, is now indistinguishable from the resistome of the treated groups, suggesting that the untreated animals will, with some delay, eventually acquire the resistome of the treated animals they share a pen with.

Fig. 4

The effect of treatment group on the resistome as shown as by nMDS at each sampling week. Significance of treatment group was evaluated by PERMANOVA at each time point and the described variance of treatment status is represented by the R²-value and corresponding p-value

Fig. 5

The effect of treatment group on the enteric microbiome as shown as by nMDS at each sampling week. Significance of treatment group was evaluated by PERMANOVA at each time point and the described variance of treatment status is represented by the R²-value and corresponding p-value

The microbiome is less affected by antibiotics than the resistome

Having determined the temporal dynamics of resistome transfer, we next hypothesized that a similar effect was present in the taxonomic composition of the enteric microbiome. To address this, we performed 16 S rRNA gene sequencing on the same set of samples and subjected this metataxonomic data to the same multivariate analysis as we did with the ARG-data. First, we observed a large and expected effect of time on the microbiome at 21.7% of explained variance across all samples. In contrast to the resistome, however, only 0.73% of the variance could be attributed to treatment group, although a more substantial and significant effect was observed from the interaction of group and time (R² = 2.18%, p = 1.5 × 10^− 5). Given this interactive effect, we again analysed the metataxonomic data within individual weeks, but observed little effect on the composition of microbiome as a function of treatment group.

Since multivariate analysis failed to show any difference between the microbiomes of Mix-treated and Mix-untreated, we instead investigated whether or not any individual genera might be differentially abundant between these groups. Only a single genus, the Prevotellaceae NK3B31 group, was systematically increased in abundance in the Mix-treated group, highlighting that the microbiome between groups Mix-treated and Mix-untreated was indeed very similar.

Discussion

In this study, we investigated the transfer of ARGs and bacteria between antibiotic treated pigs and their untreated pen mates. Our results clearly suggest that such transfer occurs and that physical separation is needed to avoid transfer of resistant bacteria and/or resistance genes to untreated pigs.

Our observations are in line with previous studies, which have also found an increase in ARGs after in-feed antibiotic treatment [19], and especially noteworthy is how the temporal dynamics have been found by others: Ghanbari et al. used shotgun sequencing to show an increase followed by a decrease of resistance genes after orally administered oxytetracycline [9], an effect lasting at least 21 days which is well in agreement with our data (Fig. 2a) as well as our prior study [16]. Although intramuscular injection of antibiotics would appear to have little effect in the gut, it has been demonstrated that tetracycline injection results in prolonged concentrations in both plasma and faeces, detectable even after 14 days although initial concentrations in plasma were much higher than in faeces [20]. Intramuscular administration of tetracycline provides considerably higher plasma concentration and considerably lower intestinal concentrations of tetracycline than peroral administration although the bioavailability of different tetracyclines differ considerably. Therapeutic concentrations of tetracycline against enterotoxigenic E. coli cannot be expected after parenteral treatment, whereas a general effect on the intestinal microbiome and resistome is evident [24,25,26,27]. Moreover, another study has found that intramuscular administered oxytetracycline in the weaning unit resulted in an elevated number of ARGs measured 2 days after the end of a 5-day tetracycline treatment followed by a reduction at the end of weaning period [21]. Apart from the mere elevation of ARGs in treated animals, a key result was the transfer of a treated phenotype from treated to untreated animals when these were reared in the same pen. Apart from our recent analysis this phenomena in an industrial farm [16], quantitative assessments of such transfer in realistic models remain scarce. Recently, however, a study very similar to ours – although using cockroaches - did indeed confirm our fundamental conclusions, namely that a treated phenotype is actively transferred to an untreated one [28].

Interestingly, the Untreated group had a slight uptick in ARG-load in weeks 4 and 5, even though this group had not received any antibiotics nor been in physical contact with treated pigs (Fig. 2b). A potential explanation could be operator transfer or air-borne transfer of bacteria bound to dust particles as found in other studies [29, 30]. An important consideration from this data is also the general decrease in resistance as well as the apparent convergence of resistance load for all groups, suggesting that the load of ARGs in the final pork product is both low and independent of the treatment status of the animal.

ARGs have been found in other studies to rise in abundance after antibiotic treatment, such as in Græsbøll et al. (2017) where the abundance of tet(A), tet(B), tet(O) and tet(W) genes increased after treatment with oxytetracycline in weaning pigs [24]. In alignment with their data, we found a similar pattern of resistance genes, namely six tetracycline resistance genes which where elevated as a response to treatment. Curiously, two aminoglycoside resistance genes (aadA1 and aphA3) as well as a streptothricin resistance gene (sat4) were found to be significantly higher in treated pigs even though they are not associated with tetracycline resistance, though they might be co-selected for as observed in a Salmonella strain isolated from pigs containing both aadA1 and tet(A) [25]. However, aadA1 was found to be significantly different between the groups in week 10 and 13 (7 and 10 weeks after treatment) at the same sampling points as tet [32]. aphA3 and sat4 were significantly different in abundance at week 3 (during antibiotic treatment) Interestingly, aphA3 and sat4 is found on the same ICE-like element in Erysipelothrix rhusiopathiae also harbouring tet(M) in the genome isolated from a pig farm [26]. In contrast, Andersen et al. (2023) observed co-selection for usage of all antibiotics, except aminoglycosides [31], and Johnson et al. (2016) had similar observations, namely that resistance genes are co-selected for owing to the fact that they are usually found in genetic clusters [32]. A significant effect of tetracycline treatment was also observed in the present study for the macrolide resistance gene ermF (Fig. 3), although the curves do not show any clear pattern and the relationship therefore appears somewhat unpredictable. However, in contrast, Birkegård et al. (2017) found a negative correlation between tetracycline treatment and the abundance of another macrolide resistance gene ermB, while macrolide treatment was positively correlated with the abundance of ermF [33]. Such seemingly contradictory observations underline the complexity of the relations between exposure to antibiotics and abundance of resistance genes. Of further interest is the scale of these values, namely how genes for tetracyclin resistance approach 0.3 copies per 16 S gene, whereas the gene for erythromycin resistance at most reaches values of 0.002.

In contrast to the resistome, treatment status had a smaller effect on the composition of the microbiome. Although the Treated and Untreated groups had distinct microbiomes from week 4 to the end of the study, the pigs in the Mixed group, having shared the same pen, appeared entirely similar. The microbiome of treated and untreated pigs has been studied in a recent study, in which intramuscular oxytetracyline resulted in a shift of the faecal microbiome lasting at least 14 days after treatment [27]. Of more interest is the lack of separation in the microbiomes of the pigs in the mixed group, which suggest that sharing a pen has a stronger effect than antibiotic intervention. Moreover, the lack of separation is in stark contrast to the resistome data, in which Mix-treated and Mix-untreated where strongly separated at several time-points. Since bacterial composition is conserved while resistome is not, a valid explanation might be that specific bacteria accumulate resistance genes through acquisition and proliferation of plasmids. Alternatively, resistance genes may be carried by a diverse set of low abundant bacteria which would not be picked up in our analysis.

The detailed analysis in our experimental design only allowed for one pen per treatment, which limits the generalization of our observations. However, this controlled setting - as well as the particularity of our analysis - allowed us to clearly quantify the effects of antibiotic treatment in pigs. Importantly, these observations are in line with what is seen in farm trials [16]. We also note that our observations are based on controlled injection of a single antibiotic – tetracycline – and may not hold true for different antimicrobials, although we believe that the overall pattern is likely to be general.

Conclusion

In conclusion, this study shows clear evidence of transfer of resistance genes from antibiotic treated pigs to untreated pen mates, suggesting that treated and untreated animals should be separated to minimize overall resistance levels. Regardless of treatment status, however, resistance levels decrease in all pigs as they age, and as they approach slaughter, no appreciable difference can be observed between treated and untreated pigs, suggesting that treatment regimen has little consequence for the end consumer. Despite this, an important conclusion from our results is that the enterotype of the treated pig, namely high levels of resistance genes, is readily transferred to untreated animals who will correspondingly contribute to the overall load of resistance genes in a given farm.