Mobility of β-lactam resistance under ampicillin treatment in gut microbiota suffering from pre-disturbance

Ingestion of food- or waterborne antibiotic-resistant bacteria may lead to dissemination of antibiotic resistance genes (ARGs) in the gut microbiota. The gut microbiota often suffers from various disturbances. It is not clear whether and how disturbed microbiota may affect ARG mobility under antibiotic treatments. For proof of concept, in the presence or absence of streptomycin pre-treatment, mice were inoculated orally with a β-lactam-susceptible Salmonella enterica serovar Heidelberg clinical isolate (recipient) and a β-lactam resistant Escherichia coli O80:H26 isolate (donor) carrying a blaCMY-2 gene on an IncI2 plasmid. Immediately following inoculation, mice were treated with or without ampicillin in drinking water for 7 days. Faeces were sampled, donor, recipient and transconjugant were enumerated, blaCMY-2 abundance was determined by quantitative PCR, faecal microbial community composition was determined by 16S rRNA amplicon sequencing and cecal samples were observed histologically for evidence of inflammation. In faeces of mice that received streptomycin pre-treatment, the donor abundance remained high, and the abundance of S. Heidelberg transconjugant and the relative abundance of Enterobacteriaceae increased significantly during the ampicillin treatment. Co-blooming of the donor, transconjugant and commensal Enterobacteriaceae in the inflamed intestine promoted significantly (P<0.05) higher and possibly wider dissemination of the blaCMY-2 gene in the gut microbiota of mice that received the combination of streptomycin pre-treatment and ampicillin treatment (Str–Amp) compared to the other mice. Following cessation of the ampicillin treatment, faecal shedding of S. Heidelberg transconjugant persisted much longer from mice in the Str–Amp group compared to the other mice. In addition, only mice in the Str–Amp group shed a commensal E. coli O2:H6 transconjugant, which carries three copies of the blaCMY-2 gene, one on the IncI2 plasmid and two on the chromosome. The findings highlight the significance of pre-existing gut microbiota for ARG dissemination and persistence during and following antibiotic treatments of infectious diseases.


INTRODUCTION
Antimicrobial resistance (AMR; here specifically limited to antibiotics, i.e. antibacterial agents) is a serious public health issue threatening the effective prevention and treatment of an ever-increasing range of bacterial infections [1]. Long-term extensive use of antimicrobials has led to the high prevalence of antibiotic-resistant bacteria (ARB) in clinical settings, agriculture systems and the environment [2][3][4][5]. A One Health approach has been taken globally, including reduction of antimicrobial use in human medicine and agriculture, in order to reduce AMR transmission to humans via the environment and food consumption [6][7][8][9]. To evaluate the potential health impacts of exposure to food-or waterborne contamination with ARB, it is critical to understand the dynamics of ingested ARB and the antibiotic-resistance genes (ARGs) that they carry in the host OPEN ACCESS gut microbiome, and how this varies according to factors such as treatment with antibiotics.
Firmicutes and Bacteroidetes are the dominant phyla in a healthy gut microbiota [10]. These bacteria produce short-chain fatty acids that maintain a mildly acidic gut environment against the colonization of opportunistic Enterobacteriaceae [11,12]. Members of the gut microbiota also compete with the intruding bacteria for niches and nutrients [13], and educate and maintain the host immune system to generate rapid and efficient response against intruders [14,15]. Thus, a healthy gut microbiota may resist the colonization of ARB following ingestion and therefore reduce the opportunity for horizontal transfer of plasmid-borne ARGs.
Plasmid conjugation is an efficient means for dissemination of ARGs from ARB to commensal and pathogen bacteria in the gut microbiota [16,17]. In general, antibiotics cause dysbiosis, thereby enabling ARB colonization and promoting conjugative ARG dissemination in the gut microbiome [18]. However, the results from modelling and in vitro experiments [19] show that antibiotics have the potential to promote or suppress conjugation through selection or inhibition of the donor, recipient or transconjugant. In the pre-existing normal mouse gut microbiota, antibiotics also have selective and suppressive effects on donor and recipient that enhance or inhibit conjugation of ARG-bearing plasmids during antibiotic treatments [20]. Since the gut microbiota often suffers from disturbance by various factors, such as diet, bacterial infection and antibiotic treatment, it is important to understand the effects of antibiotics on conjugative transfer of ARGs in microbiota with pre-existing disturbance. As a proof of concept study, we explored the impact of ampicillin on mobility of βlactam resistance in the gut microbiota that had suffered predisturbance by streptomycin.
In a previous study, co-infection of mice with multiple βlactamresistant bacterial strains favoured colonization of Escherichia coli O80:H26 and enabled conjugative transfer of its bla CMY-2 gene via an IncI2 plasmid under ampicillin treatment [21]. Building on this model, we used the E. coli O80:H26 strain as a donor, along with a βlactam-susceptible Salmonella Heidelberg, as a recipient in the present study. To elucidate how antibiotics may influence conjugation in the gut microbiota, we compared the dynamics of the donor, recipient, transconjugant and βlactam resistance genes in mice that received ampicillin treatment, streptomycin pre-treatment, a combination of streptomycin pre-treatment and ampicillin treatment, or no antibiotics as control.

Bacteria
E. coli O80:H26 (EC-107) is a multi-antibiotic-resistant strain isolated from a chicken farm in Ontario, Canada. E. coli O80:H26 carries five plasmids: IncI2, IncY, IncFII, ColRNAI and a plasmid with no detectable Inc type [21]. IncI2 is a conjugative plasmid encoding a bla CMY-2 gene and IncY is a mobilizable plasmid encoding a bla TEM-1B gene. In vitro and in vivo transfer of the IncI2 but not IncY plasmid was detected using E. coli O80:H26 as donor and E. coli O16:H48 as recipient in the previous study [21].
S. enterica serotype Heidelberg (12-6342) is a human clinical isolate that does not carry any bla gene and is susceptible to βlactam antibiotics [22]. S. Heidelberg carries two plasmids: IncX1 and ColRNAI. In the present study, S. Heidelberg was used as a recipient of βlactam resistance. To facilitate recovery of S. Heidelberg, a spontaneous rifampicin-resistant mutant was generated. In brief, S. Heidelberg was cultured overnight in Luria-Bertani (LB; Miller formulation, Difco, Fisher Scientific, Ottawa, ON, Canada) broth at 37 °C. A 1.0 ml overnight culture was pelleted, resuspended in 100 µl LB broth and spread on LB agar supplemented with 50 µg ml −1 rifampicin (LB-R). After 24 h incubation, resistant colonies were selected and sub-cultured on LB-R agar 20 times to generate and maintain a S. Heidelberg rifampicin-resistant mutant culture.

In vivo conjugation
Experiments and procedures involving mice conformed to guidelines established by the Animal Care Committee at the

Impact Statement
Plasmid conjugation is an effective means for bacterial dissemination of antibiotic-resistance genes (ARGs) in the gut microbiota. Early mouse studies showed conjugative transfer of ARGs in the gut under positive antibiotic selection pressure. Recent studies demonstrated ARG transfer in the absence of antibiotic selection pressure in mice with pre-diminished gut microbiota. This study was the first to explore the impacts of interaction between antibiotic selection pressure and pre-existing gut microbiota on the dynamics of conjugative transfer of ARGs. Our findings showed that the combination, compared to either one of the two factors, positive antibiotic selection pressure and pre-existing gut dysbiosis, promoted significantly higher and possibly wider dissemination of ARGs and prolonged the persistence of ARGs in the gut microbiota. This study points to a new direction for exploring pre-existing gut microbiota for better elucidation of the mechanisms of conjugative transfer of ARGs during antibiotic treatments of infectious bacterial diseases.
Ottawa Laboratory-Fallowfield, Canadian Food Inspection Agency. 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 or antibiotic treatment, and then housed three or four per cage (Optimice, Animal Care Systems, CO, USA) with water and feed was provided ad libitum. A total of 68 mice were randomly assigned into 2 sets of 4 groups (a total of 8 groups) to investigate the shedding of the donor and/or recipient bacteria and the transfer of plasmids carrying βlactam resistance genes under various antibiotic treatments (Table 1). One set of mice were inoculated with only the recipient bacteria and the other set were inoculated with the recipient followed by the donor bacteria 1 h later. Bacterial inocula (100 µl) prepared from log-phase culture containing ~3.0×10 8 colony-forming units (c.f.u.) of the recipient or donor bacteria in buffered peptone water (Difco) were administrated via oral gavage. Four different treatments were tested in this study: (1) ampicillin treatment (Amp), provided immediately following bacterial inoculation via drinking water (0.16 mg ml −1 , equivalent to 30 mg ampicillin kg −1 of body weight per day) ad libitum for 7 days; (2) streptomycin pre-treatment (Str), provided once via oral gavage (20 mg per mouse) 24 h before bacterial inoculation; (3) a combination of streptomycin pre-treatment and ampicillin treatment (Str-Amp); and (4) a control without the use of antibiotics (Ctl). Each treatment was applied to a group of mice from each set, one with only the recipient inoculation and the other with both the recipient and the donor inoculation. Fig. 1 shows the schedule of procedures for the group of mice which received inoculation of both S. Heidelberg and E. coli O80:H26, and treatments with both streptomycin and ampicilln. Faecal pellets were collected from all mice on −3 (baseline), 0 (bacterial inoculation), 1, 2, 3, 7, 14, 21 and 42 day post-infection (p.i.). Pellets were processed as described by Laskey et al. [21] for DNA extraction and bacterial enumeration. The three selective agars, CHR-F, XLT4-R and XLT4-FR, were used to enumerate the donor, recipient *Ampicillin treatment was provided via drinking water (0.16 mg ml −1 ) immediately following bacterial inoculation for 7 days, streptomycin treatment was provided via oral gavage (20 mg per mouse) once 24 h before bacterial inoculation, streptomycin followed by ampicillin treatment was the sequential combination of streptomycin and ampicillin treatment. †n is the number of mice used in the experiment, and the number in parentheses represents the number of mice that were euthanized on 7 day post-infection for collection of cecum tissues for histological analysis. Amp, Ampicilin; Ctl, control; EC, Escherichia coli O80:H26; SH, Salmonella Heidelberg; Str, streptomycin. and putative transconjugant bacteria, respectively, with a detection limit of 2.2 log 10 c.f.u. g −1 in faeces. At 7 days p.i., some of the mice inoculated with both recipient and donor were euthanized, consisting of five, five, six and four mice from the groups with the Amp, Str and Str-Amp treatment and the control, respectively. Tissue specimens of the cecum were collected from these mice and immediately stored in 10 % neutral buffered formalin for histological examinations.

Whole genome sequencing
Putative transconjugant bacteria were whole-genome sequenced and sequence data were analysed using the MOB-suite software tool v2.1.0 [23,24]. Representative putative transconjugant colonies isolated from selective agar plates (up to five colonies per time point) were subjected to genomic DNA extraction as described by Laskey et al. [21]. Whole-genome sequencing was performed using an Illumina MiSeq system and/or an Oxford Nanopore MinION sequencer (Oxford Nanopore, Cambridge, MA, USA) at the National Microbiology Laboratory (Guelph, ON, Canada). All short-and long-read data were deposited to the National Center for Biotechnology Information (NCBI) under BioProject PRJNA674061. The raw reads along with assemblies of genomes and plasmids were deposited under the BioSample accession numbers (SAMNs) listed in Table S1 (available in the online version of this article). Illumina raw reads were assembled using the shovill v1.1.0 pipeline (https:// github. com/ tseemann/ shovill) with the following parameters: --gsize 5000000 --assembler spades --trim --depth 0 --mincov 0 --minlen 0. Hybrid assemblies utilizing Nanopore and Illumina raw reads were assembled using unicycler v0.4.7 run under default parameters. All assemblies were manually reviewed to confirm the completeness of the chromosome and any plasmids present. As part of the validation process, complete plasmid assemblies were mapped against raw reads using the Snippy [25] pipeline to assess coverage and any potential coverage gaps. The assembled sequences were further analysed using the MOB-suite v2.1.0. and Prokka [26] software tools. An IncI1 plasmid map was rendered using the UGENE software [27] and the plasmid was annotated using Prokka v1.13.3. A gene map in a chromosomal range of an E. coli O2:H6 transconjugant was rendered using the DNA Features Viewer Python library (https:// github. com/ Edinburgh-Genome-Foundry/ DnaFeaturesViewer) and the partial genome was annotated using Prokka v1.13.3.

16S rRNA gene amplicon sequencing
DNA extracted from mouse faecal pellets was subjected to 16S rRNA gene amplicon sequencing as described by Laskey et al. [21] at the Ottawa Laboratory-Fallowfield, Canadian Food Inspection Agency (Ottawa, ON, Canada). In brief, the V3-V4 region of the 16S ribosomal RNA gene was amplified through PCR [28]. Libraries were prepared and sequenced using a MiSeq system (Illumina). Raw read data was demultiplexed and then analysed using Qiime2 [29] through a modified version of the Qiime2 pipeline created by Forrest Dusseault (https:// github. com/ forestdussault/ AmpliconPipeline). Data analysis and visualization were performed using the R package and GraphPad Prism 8.0 software (San Diego, CA, USA).

Statistical analysis
Differences in the conjugation frequency, mean abundance of each target bacterium and relative abundance of each phylum or family in the 16S rRNA gene community profiles between the treatment groups on the same sampling day were analysed with Brown-Forsythe and Welch analysis of variance (ANOVA) tests. Differences in the inflammatory score between the treatment groups were analysed using the Kruskal-Wallis test. All correlations were tested using the Pearson correlation test. The treatment groups contained up to 12 mice (Table 1), and a mean value derived from technical replicates from one faecal pellet of each mouse on each sampling date represents one datum point. Data were analysed using GraphPad Prism 8.0 software. A P value <0.05 was considered statistically significant.  (Figs 2a-d and S1). In addition, S. Heidelberg was shed in faeces for a longer period of time following co-infection with the donor and recipient than mono-infection with the recipient under all antibiotic treatments (Fig. 2b-d and f). Furthermore, the abundance of S. Heidelberg was significantly higher in faeces of mice co-infected with the donor and recipient than those mono-infected with the recipient from 2 to 42 days p.i. under Str-Amp treatment (Figs 2d, f and S2d).

Horizontal transfer of conjugative plasmids
To confirm horizontal transfer of βlactam resistance, putative transconjugants were subjected to whole-genome sequencing analysis and plasmid characterization with the MOB-suite tool v2.  (Fig. S3). Most of the identified genes on the IncI1 plasmid were of E. coli origin, suggesting a stable long-lived plasmid residence in E. coli. These genes are related to stress response, such as SOS response, toxin-antitoxin system and plasmid mobility, and could possibly contribute to conjugative transfer of the IncI1 plasmid. Analysis of the E.coli O2:H6 complete genome (NCBI BioSample SAMN16634233) identified two copies of the bla CMY-2 gene on the chromosome and one copy on the IncI2 plasmid. Both copies of the bla CMY-2 gene on the chromosome are adjacent to transposase ISEcp1 (Fig. 3), suggesting a possible movement of the bla CMY-2 gene from the IncI2 plasmid to the chromosome.

Dynamics of the β-lactam resistance genes
The E. coli O80:H26 donor carries one copy of the IncI2 plasmid encoding one copy of the bla CMY-2 gene and one copy of the IncY plasmid encoding one copy of the bla TEM-1B gene. The abundance of both genes was determined by qPCRs for investigating their transmission dynamics. Neither of the two genes was detected at a detection limit of 4.0 log 10 copies g −1 of faeces from mice mono-infected with the S. Heidelberg recipient (data not shown). From mice co-infected with the donor and recipient, both genes were detected for only 1 day in the control group, as the donor bacteria passed transiently through the mouse gut (Fig. 4a) Fig. S4b). The dynamics of bla CMY-2 and bla TEM-1B were similar in the Amp group (Fig. 4b). In comparison, bla CMY-2 abundance remained high while bla TEM-1B abundance decreased in the Str-Amp group after cessation of the ampicillin treatment (Fig. 4d). The ratio of bla CMY-2 to bla TEM-1B was significantly (P<0.05) higher from 2 to 42 days p.i. in the Str-Amp group compared to the Amp group (Fig. 4e, f).

Correlations between gut microbiota and transmission of β-lactam resistance genes
In order to determine correlations between gut microbiota and transmission of the βlactam resistance genes, the taxonomic composition of gut microbial communities of mice co-infected with the donor and recipient bacteria was further analysed using 16S rRNA gene amplicon sequencing.
In the control group, the composition of gut microbiota was  The following genes were identified in the vicinity of the bla CMY2 gene: sugE, quaternary ammonium compound resistance protein; blc, outer-membrane lipoprotein; xerC, tyrosine recombinase; malE, maltose-binding periplasmic protein; malF, maltose transport system permease protein; malG, maltose transport system permease protein; psiE, phosphate starvation-inducible membrane protein. relatively stable. The microbial community was dominated by Firmicutes, mainly the families Ruminoccoccaceae and Lachnospiraceae, and Bacteroidetes, mainly the families Lactobacillaceae and Bacteroidaceae (Fig. 5a, b). In the Amp group, the relative abundance of Proteobacteria, mainly the family Enterobacteriaceae, increased during the treatment from 0 to 7 days p.i., decreased after the cessation of ampicillin treatment and at 42 days p.i. returned to normal, a range that was not significantly (P>0.05) different from the control (Tables S3 and  S4). In the Str group, the relative abundance of Proteobacteria, mainly the family Enterobacteriaceae, increased on 0, 1 and 2 days p.i. and returned to normal at 7 days p.i. In the Str-Amp group, the relative abundance of Proteobacteria, mainly the family Enterobacteriaceae, increased from 0 to 7 days p.i. and returned to normal at 42 days p.i. (Fig. 5a, b). Expansion of Escherichia-Shigella and Salmonella relative abundance contributed to the increase of Enterobacteriaceae relative abundance (Fig. S5). The gut microbial diversity was reduced by the streptomycin pre-treatment and/or ampicillin treatment (Fig. S6). After treatment cessation, the gut microbiota gradually returned towards the original balance (Figs S6 and S7). The correlogram shows that the abundance of tranconjugant and bla CMY-2 gene is positively correlated with the relative abundance of Proteobacteria and Enterobacteriaceae and negatively with Firmicutes, which are unlikely to act as recipients (Fig. 5c). The abundance of E. coli O80:H26 and bla TEM-1B gene are negatively correlated with the relative abundance of Firmicutes, Ruminoccoccaceae and Lachnospiraceae, and the abundance of S. Heidelberg is also negatively correlated with the relative abundance of Firmicutes.

Inflammation in the mouse gut
Cecum specimens were collected at 7 days p.i. from mice co-infected with the E. coli O80:H26 donor and S. Heidelberg recipient for histopathological analysis. Inflammation was observed in cecum tissue specimens from 100, 20 and 100 % of mice in the Amp, Str and Str-Amp groups, respectively (Fig. 6a). No inflammation was found in the control mice. Fig. 6(b, c) shows the inflamed and normal cecum tissues. The inflammatory score of the Str-Amp group was significantly [H (3)=16.90, P<0.001] higher than that of the Str or control groups (Fig. 6a). These scores seemed to be positively associated with the relative abundance of Proteobacteria at 7 days p.i. Specifically, the mean inflammatory scores were 3.7, 2.6, 0.4 and 0, and the corresponding mean relative abundance of Proteobacteria was 0.8391, 0.4852, 0.0012 and 0.0002 for the Str-Amp, Amp, Str and control groups, respectively (Fig. 5a).

DISCUSSION
To explore the impact of antibiotics on ARG mobility in pre-disturbed gut microbiota, mice were subjected to streptomycin pre-treatment. The pre-treatment decreased the relative abundance of Firmicutes, more specifically Ruminoccoccaceae and Lachnospiraceae, the short-chain fatty acid producers. Thereby, the treatment would likely have lowered short-chain fatty acid concentrations, increased the luminal pH of the intestine and favoured the colonization of opportunistic pathogens [10,14], such as the E. coli O80:H26 donor and the S. Heidelberg recipient. Colonization of the donor and recipient bacteria provided a base for bacterial cell-cell contact in the gut and facilitated conjugative transfer of the bla CMY-2 gene via the IncI2 plasmid in the Str group. In comparison, in the Str-Amp group, despite benefitting the E. coli O80:H26 donor in reaching high abundance, ampicillin might kill the actively growing βlactam-susceptible S. Heidelberg at the initial stage of infection in the pre-disturbed gut microbiota. Thus, the limited abundance of recipient might lead to a lower conjugation frequency in the Str-Amp group compared to the Str group at 1 day p.i. In support of our findings, Lopatkin et al. [19] demonstrated with mathematical models and in vitro bacterial culture that antibiotics may reduce conjugation frequency by reducing the sizes of either or both of the parental populations. Hall et al. [35] also reported that positive selection for plasmid-encoded traits reduced plasmid conjugation frequency in soil bacterial communities. Here, the abundance of S. Heidelberg transconjugants in the Str group increased from 1 to 3 days p.i. and then decreased at 7 days p.i., while that in the Str-Amp group increased from 1 to 7 days p.i. The different dynamics suggest that carrying the IncI2 plasmid to S. Heidelberg is a cost in the absence but a benefit in the presence of ampicillin selection pressure. In addition, the relative abundance of commensal Enterobacteriaceae expanded significantly during ampicillin treatment. The simultaneous blooming of the donor, transconjugant and commensal Enterobacteriaceae in the severely inflamed mouse gut possibly built a strong base for gene transfer among these bacteria. In support of our suggestion, Stecher et al. [17] reported that parallel blooms of S. Typhimurium and mouse commensal E. coli boosted conjugative transfer of a colicin plasmid p2 from S. Typhimurium to E. coli. In our study, the bla CMY-2 gene was transferred via the IncI2 plasmid to a mouse commensal E. coli O2:H6 strain and incorporated into two locations of its chromosome (Fig. 3), likely through an ISEcp1-mediated transposition [36]. According to the study by Hall et al. [35], such physical movement and duplication of genes between plasmid and chromosome is a common way for bacteria to acquire antibiotic resistance. Furthermore, identical IncI1 plasmids carrying no ARGs were found in both the E. coli O2:H6 and S. Heidelberg transconjugants, suggesting possible transfer of the plasmid from E. coli O2:H6 or other commensal bacteria to S. Heidelberg. Supporting our findings on complex conjugation among bacteria in the Str-Amp group, Conlan et al. [37] reported the dissemination of the carbapenemase gene to multiple bacterial species in a patient during transplant-associated multi-course antibiotic therapies. In the present study, in pre-existing normal microbiota, ampicillin treatment facilitated the co-infection of and conjugation between the donor and recipient. However, the abundance of the donor, recipient and transconjugant were much lower in the Amp group compared to those in the Str-Amp group during the entire study, except that at 1 day p.i. S. Heidelberg abundance was slightly higher in the Amp group than the Str-Amp group. The ampicillin treatment alone without bacterial infection could disturb pre-existing normal gut microbiota and reduce Firmicutes relative abundance, as shown in our previous study [21]. However, compared to the Str-Amp group, the disturbance in the Amp group was smaller, and the less disturbed microbiota might provide greater colonization resistance to the introduced bacteria and limit their growth at lower abundance. Hence, more S. Heidelberg might become dormant at greater colonization resistance in the Amp than Str-Amp group on 1 day p.i., and survive the ampicillin treatment under the protection of βlactamase producing donor [38,39]. Yet, low abundance of the donor limited the conjugation frequency in the Amp group. In addition, the donor might depend on co-infection with the recipient in establishing colonization herein, as the donor alone failed to colonize the mouse gut under ampicillin treatment in our previous study [21]. Following cessation of ampicillin treatment or removal of selection pressure, the gut microbiota gradually recovered and diminished the introduced bacteria. The more disturbed gut microbiota likely favoured longer persistence of the introduced bacteria in the Str-Amp group compared to other groups. Moreover, under the Str-Amp treatment S. Heidelberg reached significantly higher abundance and persisted much longer in mouse faeces following co-infection with E. coli O80:H26 compared to S. Heidelberg mono-infection. The findings suggest that antibiotic-susceptible opportunistic pathogens may exploit conjugative transfer of ARGs to propagate and persist in an otherwise hostile environment. Overall, pre-disturbed gut microbiota might promote high-abundance colonization of resistant bacteria under positive antibiotic selection pressure and encourage bacterial conjugation and spread of ARGs.
The E. coli O80:H26 donor carries one copy of the bla CMY-2 gene and one copy of the bla TEM-1B gene. The abundance dynamics of the bla TEM-1B gene and the donor was highly correlated, suggesting that the bla TEM-1B gene might be transmitted clonally along with the donor. Supporting this suggestion, transconjugant that carries the bla TEM-1B gene was not recovered in this and previous studies [21]. Using the ratio of bla CMY-2 to bla TEM-1B as an indicator for dissemination of the bla CMY-2 gene, the significantly higher ratio of the two genes suggested more efficient dissemination of the bla CMY-2 gene in the Str-Amp group than in the Amp group during the entire study, except 1 day p.i. The high ratios might be attributed to the transfer of the bla CMY-2 gene to S. Heidelberg and commensal E. coli, the incorporation of the bla CMY-2 gene into the commensal E. coli chromosome and possible dissemination of the bla CMY-2 gene in the gut microbiota. Furthermore, following cessation of the ampicillin treatment, the abundance of the bla CMY-2 gene remained high, even though the abundance of the donor and transconjugant decreased significantly in the Str-Amp group, suggesting possible persistence of the bla CMY-2 gene in the gut microbiota at no selective advantage.

CONCLUSION
In this study, pre-disturbed gut microbiota promoted conjugative transfer of the bla CMY-2 gene from the E. coli O80:H26 donor to S. Heidelberg, commensal E. coli and possibly other commensal Enterobacteriaceae under positive ampicillin selection pressure. Following cessation of ampicillin treatment, shedding of the S. Heidelberg and E. coli transconjugants persisted over 35 days. These findings underline the importance of pre-existing gut microbiota on dissemination of ARGs during antibiotic treatments of bacterial infection.