Escherichia coli O80:H26 (EC-107), Salmonella enterica Bredeney (SA20114778WT), and Salmonella enterica Heidelberg (SL-312) are multi-antibiotic resistant bacteria isolated from chicken and turkey farms (Table 1 and Supplementary Table S1). Individual strains or a mixture of the three bacteria were used as donors in the mouse experiments described below. The E. coli CV601gfp (O16:H48) strain carrying a green fluorescent protein gene in its chromosome was used as a recipient (Heuer et al., 2002).

Each donor strain carries three to five conjugative and mobilizable plasmids according to whole genome sequencing analysis and plasmid characterization with the MOB-suite tool (Table 1 and Supplementary Table S2). The resistance genes encoded on the plasmids may be responsible for the multi-antibiotic resistance of the bacteria. Three conjugative plasmids that encode β-lactamases were used as targets for evaluation of horizontal transfer. These plasmids were designated by their incompatibility type according to the replication initiation protein gene: an IncI2 plasmid (MGE-644) carried by E. coli O80:H26, an IncN plasmid (MGE-934) carried by S. Bredeney and an IncA/C2 plasmid (MGE-960) carried by S. Heidelberg (Figures 1A–C and Supplementary Table S2). All three plasmids possess an origin of transfer (oriT) and encode a relaxase or nickase, a type IV coupling protein and a set of type IV secretion system (T4SS) proteins. The IncI2 and IncN plasmids each carry one resistance gene, and the IncA/C2 plasmid carries ten resistance genes (Figures 1A–C). The β-lactamase genes are flanked by or in close proximity to transposase genes in all three plasmids.

Donor strains were maintained on Chromocult agar (EMD Millipore, Toronto, ON) supplemented with 4 μg/mL cefotaxime (CHR-FOT), and the recipient strain was maintained on Chromocult agar supplemented with 50 μg/mL rifampicin and 50 μg/mL kanamycin (CHR-RK).

The donor strains were inoculated into Luria-Bertani (LB; Miller formulation, Difco, Thermo Fisher Scientific, Ottawa, ON, Canada) broth supplemented with cefotaxime (4 μg/mL final) and incubated with shaking at 30°C overnight. The recipient strain was inoculated into LB broth supplemented with rifampicin (50 μg/mL final) and kanamycin (50 μg/mL final) and incubated with shaking at 30°C overnight. The cultures were pelleted and washed three times by centrifugation at 3,000 × g for 10 min followed by resuspension in 1/10 × LB broth. After the final centrifugation, the cells were suspended to a final OD₆₀₀ of 1.0 in 1/10 × LB broth. Conjugation was assessed as follows: to a 4.5 mL aliquot of 1/10 × LB broth, 500 μL of donor cell suspension was aseptically added, vortexed briefly, and 50 μL of recipient cells were aseptically added. The cells were vortexed briefly and incubated statically overnight at 30°C. Similar aliquots containing only donor strain or only recipient strain were also prepared and incubated under the same conditions.

After incubation, a 10-fold dilution series (10^–1 to 10^–8) of the cultures was prepared in sterile saline (0.85% NaCl, pH7.0). The donor:recipient crosses were plated in triplicate onto CHR-RK, CHR-FOT and Chromocult agar supplemented with 50 μg/mL kanamycin, 50 μg/mL rifampicin and 4 μg/mL cefotaxime (CHR-RKF) and incubated at 30°C for 48–72 h. Control cultures were plated similarly onto Chromocult agar, CHR-FOT, CHR-RK, and CHR-RKF and incubated under the same conditions. Recipient colonies were enumerated on CHR-RK (E. coli CV601gfp, indigo, green fluorescent) and donor colonies on CHR-FOT (E. coli O80:H26, indigo; S. Bredeney, turquoise; S. Heidelberg, cream), respectively. Transconjugant colonies were enumerated as indigo colonies producing green fluorescence under UV illumination (Blak-Ray B-100AP/R lamp, VWR International, Mississauga, ON, United States) on CHR-RKF. Conjugation frequencies are expressed as the ratio of enumerated transconjugants to enumerated donors.

Experiments and procedures involving mice conformed to guidelines established by the Animal Care Committee at the Canadian Food Inspection Agency, Ottawa Laboratory (Fallowfield). Female C57BL/6 mice at the age of 28 days were purchased from Charles River Laboratories (Saint Constant, QC, Canada). Mice were mixed and acclimatized for 2 weeks prior to bacterial inoculation and/or antibiotic treatment, and then housed three or four per cage (Optimice^®, Animal Care Systems, Centennial, CO, United States) with water and feed provided ad libitum. Independent experiments were carried out using various donor bacteria in the presence or absence of ampicillin treatment (Table 2) to investigate the transfer of plasmids carrying β-lactam resistance genes. Bacterial inocula (100 μL) containing ∼3.0 × 10⁸ colony forming units (CFU) of E. coli O80:H26, S. Heidelberg or Bredeney or the mixture of ∼1.0 × 10⁸ CFU of each of the three bacteria as donors was given to each mouse, followed by 100 μL of ∼3.0 × 10⁸ CFU of E. coli CV601gfp as recipients through oral gavage an hour later. Immediately following the bacterial inoculation, some mice (Table 2) were provided ampicillin in drinking water ad libidum (0.16 mg ampicillin/mL water, which is equivalent to a dosage of ∼30 mg/kg per day based on each mouse weighing 20 g and consuming an average of 5 mL water a day). Fecal pellets were collected from all mice on day 0 and 1, 2, and 5 day post infection (dpi). One set of pellets were immediately stored in dry-ice and then at −80°C for microbiome analysis and the other set kept in ice for bacterial culture. Within the collection day, pellets kept in ice were weighed and then homogenized in 1.0 ml phosphate-buffered saline (PBS, pH 7.2). The homogenates were 10-fold serially diluted in PBS and suspensions plated on Chromocult agar supplemented with antibiotics as described above to enumerate donors, recipients and putative transconjugants. Following 24 h incubation at 37°C, bacteria on plates were enumerated as described above. On 7 dpi, all mice were euthanized and tissues of small intestine, cecum and colon were collected and immediately stored in 10% neutral buffered formalin for histological examinations.

Donor, recipient and putative transconjugant bacteria were characterized by whole genome sequencing. The sequence data were analyzed using the MOB-suite software tools (Robertson and Nash, 2018). Genomic DNA was isolated from the bacteria using the automated Qiagen EZ1 DNA tissue kit, according to manufacturer’s instructions, except 180 μL of G2 buffer was used with 10 μL of proteinase K and 10 μL of lysozyme (10 mg/mL; Sigma-Aldrich, Gillingham, United Kingdom). To characterize donor and recipient bacteria, PacBio and Illumina sequencing was used. PacBio sequencing was performed at the Génome Québec Innovation Centre (McGill University, Quebec, Canada) using single-molecule real-time (SMRT) cells in an RSII sequencer, which produced 120,000 to 150,000 reads per sample, with an average read length of 11,000 bp. Illumina sequencing on MiSeq version 3 (600-cycle kit, Illumina, MS-102-3003) using Nextera XT libraries (Illumina, FC-131-1031) was performed at the National Microbiology Laboratory (Guelph, ON, Canada) to a target of 60-fold coverage. All short and long read data was deposited under NCBI SRA study number SRP219110 under BioProject PRJNA560883, respectively. The final complete assemblies of genomes and plasmids were deposited under the accession numbers listed in Table 1.

The acquisition of bla genes by putative transconjugants was confirmed as follows. Representative colonies isolated from media plates for enumerating transconjugants (up to 3 colonies per plate) were subjected to whole genome sequencing using an Illumina MiSeq system and/or an Oxford Nanopore MinION sequencer (Oxford Nanopore, Cambridge, MA, United States) at the National Microbiology Laboratory (Guelph, ON, Canada). Illumina sequencing was as described above. Oxford Nanopore sequencing was performed according to the default manufacturer protocol for rapid barcoding. Samples were prepared using either SQK-RBK001 or SQK-RBK004 rapid barcoding kits and subsequently ran on a FLO-MIN106 R9.4 flow cell. Each multiplexed run produced between 4,719 and 111,488 reads per sample, with the mean read length ranging between 3,485 and 11,880 bp. Albacore v2.1.3 (Oxford Nanopore) was used to perform demultiplexing, base-calling and quality filtering of the raw reads.

Hybrid de novo assemblies of transconjugants were produced using the Unicycler pipeline v0.4.3 (Wick et al., 2017). Hybrid de novo assemblies of the recipient and donor strains were performed HGAP v3.1 PacBio assembly pipeline. All assemblies were manually reviewed to confirm completeness of the chromosome and any plasmids present. As part of the validation process, complete plasmid assemblies were mapped against raw reads using snippy (Seemann, 2015) pipeline to assess coverage and any potential coverage gaps. The assembled sequences were further analyzed using the MOB-suite (Robertson and Nash, 2018) and Prokka (Seemann, 2014) software tools. Plasmid maps were rendered using the UGENE software (Okonechnikov et al., 2012) and the plasmids were annotated using Prokka version 1.13.3.

DNA was extracted from each fecal pellet using the NucleoSpin^® Soil DNA extraction kit (Macherey-Nagel, Germany), per the manufacturer’s protocol. Concentrations of DNA samples were determined using a combination of the Qubit 2.0 Fluorometer (Thermo Fisher Scientific, Ottawa, ON, United States) with the Qubit dsDNA HS Assay Kit, as well as the Quanti-iT dsDNA Assay Kit (Thermo Fisher Scientific) with a BioTek FLx800 microplate fluorescence reader (Thermo Fisher Scientific). DNA libraries were prepared following Illumina’s 16S Metagenomic Sequencing Library Preparation protocol (Illumina). In brief, the V3-V4 region of the 16S ribosomal RNA gene was amplified through PCR (Klindworth et al., 2013) using KAPA HiFi HotStart Ready Mix (Roche, Cape Town, ZA, United States). PCR products were subsequently purified using Agencourt AMPure XP beads (Beckman Coulter, Mississauga, ON, United States) and indexed using primers from the Nextera XT Index Kit (Illumina, FC-131-2001). Indexed PCR products were purified and then subjected to quantitation and quality verification in the QIAxcel Advanced System with a QIAxcel DNA High Resolution Kit (Qiagen). Libraries were diluted accordingly, pooled, and denatured prior to loading into the MiSeq v3 Reagent Kit (Illumina, MS-102-3003) cartridge with approximately 15% PhiX control from the PhiX Control v3 Kit (Illumina, FC-110-3001). Libraries were sequenced using a MiSeq system (Illumina) at the Canadian Food Inspection Agency Ottawa Laboratory (Fallowfield). Raw read data was demultiplexed and then analyzed using Qiime2 (Bolyen et al., 2018) through a modified version of the Qiime2 pipeline created by Forrest Dusseault¹. Denoising, filtering, and clustering of OTUs in Qiime2 was conducted using the DADA2 option. The biom file, tree file and metadata file generated from QIIME2 were combined into a phyloseq object using R package phyloseq² for further analysis and visualization.

Intestinal segments from small intestine, cecum and colon were prepared in Swiss rolls and fixed in 10% neutral buffered formalin for at least 24 h. Fixed tissues were embedded in paraffin, sectioned, and stained with hematoxylin and eosin (Fischer et al., 2008). Inflammation was quantitated by evaluating submucosal edema, PMN infiltration, goblet cell hyperplasia, and epithelial damage and given a total score from 0 – 4: 0, no disease; 1, minimal; 2, mild; 3, moderate; 4, marked as described by Erben et al. (2014).

Differences in the conjugation frequency or the relative fold change of Proteobacteria between treatment groups on the same sampling day, and differences in the relative abundance of each phylum or genus between sampling days within each treatment groups were tested using one-way ANOVA. Differences in the percentage of mice that developed inflammation in the cecum and colon between treatment groups were tested using the Fischer’s exact test. The correlation between the percentage of mice that developed inflammation and the relative fold change of the Proteobactia in the gut microbiota was tested using the Pearson correlation test. The treatment groups contain four to six mice (Table 2), and a mean value derived from technical triplicates from one fecal pellet of each mouse on each sampling date represents one data point. Data were analyzed using the GraphPad Prism 8.0 software (San Diego, CA, United States). A P-value < 0.05 was considered statistically significant.

The in vitro conjugation frequency was 7.8 × 10^–4, 5.3 × 10^–5, and 3.0 × 10^–4 between the E. coli O80:H26, S. Bredeney or S. Heidelberg donor and the E. coli CV601gfp recipient, respectively. This result indicated that the targeted β-lactam resistant plasmids were transferable, and therefore suitable for in vivo experimentation. In the presence of ampicillin treatment, shedding of the S. Bredeney and S. Heidelberg donors lasted for at least 5 days, with a maximum abundance in mouse feces at 1 day post infection (dpi) (Figures 2B,C). In comparison, shedding of the E. coli O80:H26 donor lasted for only 1 day, at 4.4 log₁₀ CFU/g (Figure 2A). However, when co-introduced with the two Salmonella donor strains, shedding of the E. coli O80:H26 lasted for at least 5 days, with a maximum abundance of 8.1 log₁₀ CFU/g at 1 dpi (Figure 2D). The recipient E. coli CV601gfp was shed for 1 or 2 days at a range of 2.6–4.1 log₁₀ CFU/g. Fewer than 2.7 log₁₀ CFU/g transconjugants were recovered, and only at 1 dpi from mice inoculated with a single donor strain (Figures 2A–C). In comparison, from mice inoculated with the mixture of all three donor strains, transconjugants were recovered at 1 and 2 dpi reaching 4.5 log₁₀ CFU/g (Figure 2D), accompanied by some non-fluorescent colonies at 2 dpi. The conjugation frequency in mice with all three donor strains was significantly (P < 0.001) higher than that with each donor strain (Figure 2E), although each mouse received equal numbers of donor bacteria. In the absence of ampicillin treatment, transconjugants were not detected based on a detection limit of 2.2 log₁₀ CFU/g. Shedding of the donors and recipient was less than 3.3 log₁₀ CFU/g and lasted for only 1 day (Figures 2A–D).

To confirm horizontal transfer of the β-lactam resistant plasmids, putative fluorescent and non-fluorescent transconjugants were subjected to whole genome sequencing analysis and plasmid characterization with the MOB-suite tool. Sequencing information on the representative transconjugants is available in BioProject PRJNA560883. According to the Mash distance analysis, plasmids in the transconjugants are identical to those in the corresponding donor bacteria, although there are a few mismatches likely due to sequencing errors (Supplementary Table S3). There were four different strains of transconjugants recovered from the mice that were inoculated with all three donor bacteria. Three strains were derived from the E. coli CV601gfp recipient and carried the IncI2, IncN and IncA/C2 plasmids, respectively. The other strain was a non-fluorescent E. coli O2:H6 carrying the IncI2 plasmid. Overall, the sequencing data confirmed the transfer of the conjugative plasmids from donor bacteria to the exogenously introduced E. coli CV601gfp and to an endogenously present E. coli O2:H6 recipient.

To determine if any of the treatments induced inflammation, intestine tissues were collected from each mouse on 7 dpi for histopathological analysis. Figures 3A1,A2,B1,B2 show the inflamed and normal cecum and colon tissues. In the presence of ampicillin treatment, inflammation was observed in the cecum of 0, 50, 60, and 100% of the mice infected by E. coli O80:H26, S. Bredeney, S. Heidelberg and the mixture of all three donor bacteria, respectively (Figure 3C1). Also, inflammation was observed in the colon of 0, 17, 60, and 100% of the mice from the corresponding treatment groups (Figure 3C2). The co-infection with all three donor bacteria resulted in significantly greater percentages of mice with inflammation in the cecum and colon compared to the infection with only E. coli O80:H26 (P < 0.05). No inflammation was found in mice receiving only ampicillin treatment but no bacterial inoculation, or only bacterial inoculation but no ampicillin treatment. In addition, no inflammation was found in the small intestine of all mice.

Fecal samples from all mice were subjected to the 16S rRNA gene amplicon sequencing analysis in order to determine if treatments promoted dysbiosis. In the absence of ampicillin treatment, the composition of gut microbiome in mice with or without bacterial inoculation was relatively stable. The microbial community was dominated by Firmicutes and Bacteroidetes (Figures 4A–E). As expected, ampicillin treatment disrupted the gut microbiome, with a significant (P < 0.05) decrease in the relative abundance of Firmicutes and a significant (P < 0.05) increase in that of Bacteroidetes and Proteobacteria occurring at 1 dpi (Figure 4A). The introduction of various donor bacteria had differential impacts on the alteration of the microbial composition and diversity induced by the ampicillin treatment (Figures 4B–E and Supplementary Figures S1, S2). Colonization of E. coli O80:H26 slightly alleviated the dysbiosis, while that of S. Bredeney, S. Heidelberg and the mixture of all three donor bacteria aggravated the dysbiosis. To illustrate the differential impacts, the alteration of Proteobacteria relative abundance at 1, 2, and 5 dpi compared to that at day 0 were expressed as fold changes within each mouse group (Figure 4F). The fold changes in the mice with E. coli O80:H26 inoculation and ampicillin treatment were comparable to or slightly lower than those in the mice without bacterial inoculation or ampicillin treatment. In contrast, the fold changes in the mice with S. Bredeney, S. Heidelberg or the mixture of all three bacterial inoculation plus ampicillin treatment were significantly (P < 0.05) higher compared to those of the mice with only ampicillin treatment but no bacterial inoculation (Figure 4F). The mean fold changes in the relative abundance of Proteobacteria at 5 dpi were 1, 58, 102, and 138 for the mice inoculated with E. coli O80:H26, S. Bredeney, S. Heidelberg and the mixture of all three bacteria plus ampicillin treatment, respectively. These changes were positively correlated with the percentage of mice that developed intestinal inflammation in the corresponding treatment groups (R² > 0.93, P < 0.01, Supplementary Figure S3). Since the donor and the recipient bacteria belong to the Escherichia-Shigella and Salmonella genera, the relative abundance of these two genera was analyzed. In the mice with all three bacterial inoculation and ampicillin treatment, the relative abundance of Escherichia-Shigella increased significantly (P < 0.05) from 0.0002 at day 0 to 0.1155, 0.1379, and 0.4464 at 1, 2, and 5 dpi, respectively (Figure 5E). However, in the mice with only E. coli O80:H26 inoculation and ampicillin treatment, the relative abundance of Escherichia-Shigella remained below 0.0027 over the course of the study (Figure 5B). Comparing data from these two treatment groups, there seemed to be an association between the relative abundance of Escherichia-Shigella and the colonization of E. coli O80:H26 and also the transfer of the IncI2 plasmid (Figures 2A,D, 5B,E). In comparison, the relative abundance of the Salmonella genus was greater than 0.0150 at 1, 2, and 5 dpi in the mice with Salmonella inoculation and ampicillin treatment (Figures 5C–E). In the absence of ampicillin treatment, the abundance of these two genera remained relatively stable and below 0.0040 in all mice (Figures 5A–E).