In Vitro Assessment of Antimicrobial Resistance Dissemination Dynamics during Multidrug-Resistant-Bacterium Invasion Events by Using a Continuous-Culture Device

Antimicrobial resistance poses a grave threat to public health and reduces the effectiveness of antimicrobial drugs in treating bacterial infections. Antimicrobial resistance is transmissible, either by horizontal gene transfer between bacteria or by vertical gene transfer following inheritance of genetic traits.

ABSTRACT

Antimicrobial-resistant pathogens display significant public health threats by causing difficulties in clinical treatment of bacterial infection. Antimicrobial resistance (AMR) is transmissible between bacteria, significantly increasing the appearance of antimicrobial-resistant pathogens and aggravating the AMR problem. In this work, the dissemination dynamics of AMR from invading multidrug-resistant (MDR) Escherichia coli to a community of pathogenic Salmonella enterica was investigated using a continuous-culture device, and the behaviors of dissemination dynamics under different levels of antibiotic stress were investigated. Three MDR E. coli invasion events were analyzed in this work: MDR E. coli-S. enterica cocolonization, MDR E. coli invasion after antibiotic treatment of S. enterica, and MDR E. coli invasion before antibiotic treatment of S. enterica. It was found that both horizontal gene transfer (HGT) and vertical gene transfer (VGT) play significant roles in AMR dissemination, although different processes contribute differently under different circumstances, that environmental levels of antibiotics promote AMR dissemination by enhancing HGT rather than leading to selective advantage for resistant bacteria, and that early invasion of MDR E. coli completely and quickly sabotages the effectiveness of antibiotic treatment. These findings contribute to understanding the drivers of AMR dissemination under different antibiotic stresses, the detrimental impact of environmental tetracycline contamination, and the danger of nosocomial presence and dissemination of MDR nonpathogens.

IMPORTANCE Antimicrobial resistance poses a grave threat to public health and reduces the effectiveness of antimicrobial drugs in treating bacterial infections. Antimicrobial resistance is transmissible, either by horizontal gene transfer between bacteria or by vertical gene transfer following inheritance of genetic traits. The dissemination dynamics and behaviors of this threat, however, have not been rigorously investigated. In this work, with a continuous-culture device, we studied antimicrobial resistance dissemination processes by simulating antimicrobial-resistant Escherichia coli invasion to a pathogenic Salmonella enterica community. Using this novel tool, we provide evidence on the drivers of antimicrobial resistance dissemination, on the detrimental impact of environmental antibiotic contamination, and on the danger of antimicrobial resistance in hospitals, even if what harbors the antimicrobial resistance is not a pathogen. This work furthers our understanding of antimicrobial resistance and its dissemination between bacteria and of antibiotic therapy, our most powerful tool against bacterial infection.

INTRODUCTION

The extensive use of antibiotics in both the medical and agricultural sectors for curing and preventing bacterial infection is now leading to the increase of global antimicrobial resistance (AMR) and the reduction of antibiotic efficacies (1). This worrisome development has raised worldwide concerns, fearing that we might go back to the pre-antibiotic era in terms of bacterial infection control, when even only a small cut in the finger could be fatal (2).

Both intrinsic and acquired resistance contribute to AMR in bacteria. A large fraction of acquired resistance is caused by the acquisition of antimicrobial resistance genes (ARGs). This is of particular importance for the mobility and transmissibility of acquired resistance between bacterial cells or even species, which in turn lead to the fast evolvement of multidrug-resistant pathogens. This is commonly done by transfer of ARGs carried by mobile genetic elements (MGEs) such as plasmids, integrative and conjugative elements (ICEs), and integrons (3). These MGEs are disseminated intercellularly via conjugation, natural transformation, or transduction, processes collectively termed horizontal (or lateral) gene transfer (HGT) (4). Among the three processes, conjugation is considered to be the most important one that represents the majority of HGT events (5).

The dissemination of AMR in microbial communities has been a hot spot for research due to its importance in AMR evolution. A large body of literature provided evidence on the dissemination of AMR in vitro using pure bacterial isolates or bacterial communities (6–8), on interspecies AMR dissemination in vivo using animal models or in humans (9–11), and on clinical cases confirming AMR dissemination in patients (12, 13). Attention has been focused on regulatory factors and pathways as they are potential targets for blocking AMR dissemination. HGT, in particular, conjugation, has been found to be regulated by a variety of chemicals such as antibiotics (9, 14–21), biocides (22, 23), dyes (24), heavy metals (25, 26), antiepileptic drugs (27), disinfectants (6), nanomaterials (28–30), hormones (31), feed supplements (32), ionic liquids (33), and fuel exhaust particles (34) and by biological pathways, including the SOS response (35, 36) and quorum sensing (37–40).

Whether, how, and to what extent do antibiotics influence the dissemination of AMR are key questions that lie in the core of scientists’ concerns. People not only encounter high concentrations of antibiotics during antibiotic therapies, they also receive low/trace levels of antibiotics from the environment as a consequence of widespread antibiotic pollution (41). In these scenarios, antibiotics may play the role of a stressor that put resistant bacteria at a selective advantage, leading to better survival rates of resistant bacteria that harbor genes responsible for AMR, which in turn leads to higher levels of vertical gene transfer (VGT) of these genes; they may also play the role of a regulatory substance that upregulates horizontal transfer machineries, leading to higher rates of HGT for genes responsible for AMR intercellularly. In other words, antibiotics may promote both VGT and HGT, albeit via different mechanisms and possibly at different antibiotic concentrations. A recent research study quantitatively compared VGT and HGT events during AMR dissemination using microfluidics, suggesting both processes contribute significantly to the transfer of resistance and both are stimulated by high levels of certain antibiotics (8). Nontherapeutic concentrations of antibiotics can also be selective: in some cases, the minimal selective concentration (MSC) can be as low as 1/230 of the MIC (42). This concentration is well below therapeutic concentrations but still several orders higher than environmental levels. It is generally accepted that low (nontherapeutic) concentrations of antibiotics stimulate HGT via processes such as the SOS response (17, 19, 43, 44), and evidence suggested that the conjugation machinery is upregulated by subinhibitory concentrations of antibiotic (19, 35). Nevertheless, the concentrations used in these investigations are still way higher (by several orders of magnitude) than what we may encounter in the environment (nanograms to low micrograms per liter). Interestingly, in a recent report by Lopatkin et al., conjugation does not appear to be stimulated by subinhibitory concentrations of antibiotics in a system where the conjugation machinery is constitutively expressed (45). The authors then went on to conclude that the contribution of antibiotic-promoted HGT in the dissemination of AMR is overestimated. Therefore, whether the trace level of antibiotics we may consume as environmental contamination may have a role in promoting HGT still remains an unanswered question.

Despite the extensive investigations on the dissemination mechanisms of AMR, very little research has been conducted on the dynamics of AMR dissemination in a setting with sustaining microbial communities that simulate clinical or environmental environments. To date, with only a few exceptions (7, 8, 46–48), nearly all the in vitro investigations on the dissemination of AMR were carried out in shake flasks or with conjugation assays, and only a few in vivo investigations were performed (49–51). The animal model and mucosal simulator of the human intestinal microbial ecosystem (M-SHIME) studies that best represent in vivo conditions mainly focused on the transferability of AMR and barrier of dissemination (CRISPR-Cas), without going into details of dissemination dynamics (46, 49, 50). This is understandably due to the complex nature of these models and the microbial communities used. Continuous-culture devices are excellent simulators of sustained microbial communities and can be used for quantitative analysis of microbial dynamics. Two previous studies used chemostats for the analysis of AMR dissemination between bacteria, showing antibiotic resistance plasmid transfer between two Escherichia coli strains and from Salmonella to E. coli in a chicken gut microbial consortium, without going into detailed AMR dissemination dynamics (7, 48).

Salmonella infection is one of the four major causes of diarrhea worldwide. It is caused by consumption of contaminated food of animal origin or transmission through the fecal-oral route and can lead to gastroenteritis, intestinal heat, and Salmonella bacteremia (52, 53). It was estimated that Salmonella infections cause 93.8 million cases of gastroenteritis and 155,000 deaths worldwide each year (54). Multidrug-resistant (MDR) Salmonella species raised concerns in recent years for the challenge in treating this infection (55). In this work, we simulated the invasion of multidrug resistance carried by commensal E. coli, a common bacterium in the gastrointestinal tract, to a Salmonella community using continuous-culture devices. The AMR dissemination process was monitored, and dissemination dynamics were assessed for scenarios where both MDR E. coli and Salmonella coappear, MDR E. coli invades after antibiotic treatment is initiated, and MDR E. coli invades prior to antibiotic treatments. AMR dissemination dynamics for tetracycline treatment with high therapeutic (40 mg/liter, adult dosage), low therapeutic (10 mg/liter, infant dosage), subinhibitory (1 mg/liter, MIC level), and reported environmental (0.002 mg/liter) concentrations were also determined and compared (56–59). To the best of our knowledge, this is the first detailed assessment of AMR invasion and dissemination dynamics in a sustained continuous microbial system.

RESULTS

Continuous-culture experiment settings.

An in vitro continuous-culture system was established for the assessment of AMR dissemination dynamics between E. coli and Salmonella enterica, both of which belong to the Enterobacteriaceae family (Fig. 1 and 2; see also Fig. S1 in the supplemental material). The conjugative transfer of the pRP4 plasmid from E. coli to S. enterica was experimentally confirmed (see Fig. S2). The standard curves of E. coli C600(pRP4) and S. enterica H9812 were determined to make sure of a donor/recipient cell count ratio of 1:10⁵ (see Fig. S3). Three experiments were performed, each mimicking an MDR E. coli invasion event: an event where MDR E. coli and S. enterica cocolonize an environment, an event where MDR E. coli invades an S. enterica community already under antibiotic pressure, and an event where antibiotic pressure or treatment takes place after MDR E. coli invades a stable S. enterica community. All three experiments simulate scenarios where MDR E. coli establishes itself in S. enterica infections for the purpose of assessing how AMR disseminates from E. coli to the infecting pathogen S. enterica under various antibiotic contamination, pressure, or treatment conditions. Analysis of average pRP4-to-cell ratios suggest antibiotic stress did not impact average pRP4 copies in MDR E. coli (Fig. 3A), but overall pRP4-to-cell ratios increased significantly under therapeutic antibiotic pressure in each event (Fig. 3B to D), confirming that AMR dissemination took place under antibiotic pressure during MDR E. coli invasion and that overall resistance levels increased for bacterial systems investigated in this work. Susceptibility testing of 33 S. enterica isolates from antibiotic-containing selective plates further confirmed that AMR dissemination did take place from MDR E. coli to S. enterica (see Table S1).

FIG 1.

Schematic illustration of continuous-culture device setup. Each glass bottle (left) provided fresh culture medium for three parallel glass fermentation vessels (right) simultaneously.

FIG 2.

Experimental design.

FIG 3.

The relative levels of pRP4 plasmid. (A) Levels of pRP4 plasmids in E. coli exposed to tetracycline. (B) Levels of pRP4 plasmids during MDR E. coli-S. enterica cocolonization event; data at 1 h indicate initial level after inoculation, and data at 60 h indicate level after culture stabilization. (C) Levels of pRP4 plasmids during MDR E. coli invasion event after antibiotic treatment; data at 48 h indicate an early (5 h) level right after invasion of E. coli, and data at 62 h indicate a level close to (but not yet) stabilization. (D) Levels of pRP4 plasmids during MDR E. coli invasion event before antibiotic treatment; data at 56 h indicate a level right after the addition of tetracycline, and data at 92 h indicate a level after culture stabilization. *, P < 0.05; **, P < 0.01; ***, P < 0.001. Error bars, standard deviations.

Drive of AMR dissemination by antibiotic selection during an E. coli-S. enterica cocolonization event.

The dissemination of AMR during an E. coli-S. enterica cocolonization event was analyzed by coinoculating pRP4-harboring E. coli and S. enterica at a ratio of 1:10⁵ under four tetracycline concentrations: no tetracycline, the environmental level of 0.002 mg/liter, the subinhibitory 1 mg/liter, and the therapeutic 10 mg/liter. No distinctive differences were observed for the total bacterial levels after the bacterial communities stabilized at all tetracycline levels, although the therapeutic level of tetracycline did lead to slower initial growth (Fig. 4A).

FIG 4.

AMR dissemination dynamics in the MDR E. coli-S. enterica cocolonization event. (A) Total bacterial levels. (B) Total E. coli levels. (C) Total S. enterica levels. (D) Tetracycline-sensitive S. enterica levels. (E) MDR S. enterica levels. (F) Proportions of each bacterium in continuous-culture devices. P values were calculated to compare experimental and control groups. *, P < 0.05; **, P < 0.01; ***, P < 0.001. Error bars, standard deviations. Concentrations indicate levels of tetracycline.

Converse growth patterns were found for E. coli and S. enterica (Fig. 4B and C). For the case of E. coli, higher levels of tetracycline led to faster growth and higher maxima (Fig. 4B). For S. enterica, subinhibitory and therapeutic levels of tetracycline led to lower growth rates (Fig. 4C). The trend for S. enterica is understandably due to inhibition of growth by high levels of tetracycline, while the trend for E. coli can be attributed to the better competitiveness of MDR E. coli in comparison to that of antibiotic-sensitive S. enterica in a tetracycline-stressed scenario. Therefore, these converse trends are a clear indication of antibiotic selection as a key factor for shaping growth characteristics in this cocolonization event.

Transfer of AMR from E. coli to S. enterica took place at approximately the same time under all tetracycline concentrations (60 h postinoculation) (Fig. 4E), suggesting HGT of pRP4 from E. coli to S. enterica is not significantly impacted by antibiotic pressure in this setting. However, the levels of MDR S. enterica increased drastically faster at the highest tetracycline level, with a higher final maximum. This again can be attributed to the better competitivity of MDR S. enterica in comparison to that of tetracycline-sensitive S. enterica under high tetracycline stress, which is confirmed by the finding that tetracycline-sensitive S. enterica has a lower final maximum with the therapeutic level of tetracycline (Fig. 4D).

From analysis of these bacterial community dynamics, we are unable to find the role of varied HGT rates in driving AMR dissemination with different tetracycline concentrations during a cocolonization event, as HGT took place almost simultaneously at all tetracycline levels. However, selection that is determined by VGT apparently drives the change of bacterial community structures, as depicted in Fig. 4F. A conclusion can therefore be made that primary driver for AMR dissemination is VGT in a cocolonization event. Also, no distinctive differences were found for AMR dissemination between the environmental level of tetracycline and the control experiment in this setting, indicating that an environmental concentration of tetracycline does not significantly impact the VGT process.

Contribution of HGT in AMR dissemination during MDR E. coli invasion to antibiotic-stressed S. enterica communities.

In a second simulated MDR E. coli invasion event, antibiotic treatment was first applied to a stabilized S. enterica community. Four concentrations of tetracycline were used in this investigation: control (0 mg/liter), environmental (0.002 mg/liter), low therapeutic (10 mg/liter), and high therapeutic (40 mg/liter) levels. Addition of both therapeutic levels of antibiotics led to immediate reduction of S. enterica levels by 3 to 4 orders of magnitude, while the environmental level of tetracycline led to negligible impact (Fig. 5A). No resistance to tetracycline developed for S. enterica up until the invasion of MDR E. coli (Fig. 5C), suggesting evolution of tetracycline resistance under antibiotic pressure is slow and had not taken place for 23 h after tetracycline application.

FIG 5.

AMR dissemination dynamics in the post-antibiotic-treatment MDR E. coli invasion event. (A) Total bacterial levels. (B) Total E. coli levels. (C) Total S. enterica levels. (D) Tetracycline-sensitive S. enterica levels. (E) MDR S. enterica levels; small graphs show the slopes and levels of MDR S. enterica evolvement. (F) Proportions of each bacterium in continuous-culture devices. P values were calculated to compare experimental and control groups. *, P < 0.05; **, P < 0.01; ***, P < 0.001. Error bars, standard deviations. Concentrations indicate levels of tetracycline.

An intriguing phenomenon was observed upon the invasion of MDR E. coli to the S. enterica community: all bacterial groups increased, in contrary to what was observed for the cocolonization experiment where VGT-determined selection took place and several bacterial groups developed at the expense of other groups (Fig. 5B to E). This finding clearly suggests that the mutual competition between bacterial groups does not play the key role in the dissemination of AMR in this invasion event. E. coli exposed to zero and environmental levels of tetracycline grew at a higher rate than those exposed to therapeutic levels of tetracycline (Fig. 5B), reflecting the pressure tetracycline places on the growth of E. coli, despite it containing an MDR plasmid. S. enterica levels started to increase right after the invasion of E. coli at therapeutic antibiotic concentrations (Fig. 5C), suggesting an increase of overall resistance. The timing of this phenomenon also suggests that the dissemination of AMR from E. coli to S. enterica instead of development of new AMR phenotypes by mutations in certain key genes is the primary reason for the increased resistance for S. enterica.

Analysis of the tetracycline resistance in S. enterica leads to another interesting finding: following invasion of MDR E. coli, S. enterica exposed to environmental and low therapeutic levels of tetracycline developed resistance faster than those exposed to zero and high therapeutic levels of tetracycline. The contents of MDR S. enterica increased faster at environmental and low therapeutic levels of tetracycline, with higher maximum levels (Fig. 5E). This difference suggests that HGT plays the major role in shaping AMR dissemination here, as MDR S. enterica apparently does not have better VGT-determined selective advantage at intermediate (environmental and low therapeutic) levels of tetracycline in comparison to that at a high (high therapeutic) level, and the primary driving factor for this unique AMR dissemination pattern here can only be the other probable major process, HGT. To confirm this, conjugation frequencies between MDR E. coli and S. enterica were determined (Fig. 6), showing that conjugation frequencies at these intermediate tetracycline levels are indeed higher than at higher or lower levels. A control experiment was performed at the same time showing that under the conjugation frequency measurement conditions, the numbers of live cells did not significantly change (Fig. 6), therefore eliminating the possibility that the different conjugation frequencies are an artifact due to different bacterial survival rates. Interestingly, this induction of conjugation by tetracycline appears to be tetracycline specific but not strain specific: other tested antibiotics did not show a similar effect, while a different E. coli strain showed the same conjugation induction by the environmental level of tetracycline (Fig. 7). All these analyses show that HGT plays the primary role in driving AMR dissemination in this post-antibiotic-treatment MDR E. coli invasion event and that HGT of AMR from E. coli to S. enterica is regulated by antibiotic concentrations. The changes of bacterial community structures as shown in Fig. 5F also confirm this finding: although MDR S. enterica should outperform sensitive S. enterica at a higher ratio at 40 mg/liter tetracycline than at 10 mg/liter tetracycline, interestingly, the proportion of MDR S. enterica at 10 mg/liter tetracycline toward the end of the experiment is much higher than at 40 mg/liter tetracycline, a phenomenon only possible when improved HGT rates at 10 mg/liter tetracycline account for differences in AMR dissemination. Of particular interest, an environmental level of tetracycline at 0.002 mg/liter increased the dissemination of AMR significantly via regulating HGT, experimentally showing that environmental tetracycline contamination can drive AMR dissemination, confirming the potential danger of even very low concentrations of tetracycline in the environment.

FIG 6.

Conjugation frequencies at different concentrations of tetracycline. The conjugation efficiencies of pRP4 between E. coli C600 and S. enterica H9812 in the absence of tetracycline (control) and in the presence of tetracycline at high therapeutic (40 mg/liter), low therapeutic (10 mg/liter), subinhibitory (1 mg/liter), and environmental (0.002 mg/liter) tetracycline concentrations were quantified. Table shows bacterial levels prior to and after 1-h tetracycline treatment. P values were calculated to compare experimental and control groups. *, P < 0.05; ***, P < 0.001. Error bars, standard deviations.

FIG 7.

Conjugation frequencies at environmental concentrations of antibiotics. (A) Conjugation efficiencies of pRP4 between E. coli C600 and S. enterica H9812 in the absence of antibiotics (control) and in the presence of antibiotics at an environmental (0.002 mg/liter) concentration. (B) Conjugation efficiencies of pRP4 between E. coli BW25113 and S. enterica H9812 in the absence of antibiotic (control) and in the presence of tetracycline at an environmental (0.002 mg/liter) concentration. *, P < 0.05. Error bars, standard deviations.

The dissemination of AMR upon MDR E. coli invasion is clear and significant in this experiment, which caused significant consequences as demonstrated by the increase of S. enterica at therapeutic tetracycline levels. However, it is still somewhat comforting that the invasion of MDR E. coli after tetracycline treatment has taken place did not lead to a full recovery of S. enterica levels. The invasion of MDR E. coli led to an increase of S. enterica by 1 to 2 orders of magnitude but was still 1 to 2 orders of magnitude lower than the levels prior to tetracycline treatment. Therefore, early antibiotic therapy is still demonstrated effective, although invasion of MDR bacteria partially reduces its effectiveness.

Early invasion of MDR E. coli sabotages effectiveness of antibiotic treatment.

The impact of early invasion of MDR E. coli to an S. enterica community prior to antibiotic treatment was investigated. Upon invasion of MDR E. coli, AMR dissemination took place from E. coli to S. enterica (Fig. 8B and E). Tetracycline treatment was applied to the system 24 h after MDR E. coli invasion at three concentrations: control (0 mg/liter), low therapeutic (10 mg/liter), and high therapeutic (40 mg/liter) levels. Addition of tetracycline led to the decrease of S. enterica and total bacteria by approximately 2 orders of magnitude (Fig. 8A and C), 1 order of magnitude smaller than antibiotic treatment prior to MDR E. coli invasion (Fig. 5A and C). However, unlike what was observed for post-antibiotic-treatment MDR E. coli invasion, the levels of both S. enterica and total bacteria quickly returned to the pretreatment level merely 24 h after treatment (Fig. 8A and C). Treatment with therapeutic levels of antibiotics led to higher levels of MDR E. coli and MDR S. enterica and lower levels of tetracycline-sensitive S. enterica (Fig. 8B to E), suggesting that antibiotic selection accounts for the change of bacterial community structures and drives AMR dissemination, eventually leading to similar levels of total bacteria but higher proportions of MDR bacteria (Fig. 8F). This is similar to the AMR dissemination dynamics observed for MDR E. coli-S. enterica cocolonization (Fig. 4). Nevertheless, the quick recovery of total bacteria and S. enterica levels after antibiotic treatment is a strong indication of the failure for antibiotic therapy, which further suggests that early MDR bacterial invasion can sabotage latter antibiotic treatment, despite the fact that only less than 0.1% of S. enterica cells were tetracycline resistant when antibiotic treatment started (Fig. 8A and C). This is in clear contrast to what was observed in this work, in which early antibiotic treatment prior to MDR E. coli invasion is still effective in controlling the levels of pathogenic S. enterica.

FIG 8.

AMR dissemination dynamics in the MDR E. coli invasion event before antibiotic treatment. (A) Total bacterial levels. (B) Total E. coli levels. (C) Total S. enterica levels. (D) Tetracycline-sensitive S. enterica levels. (E) MDR S. enterica levels. (F) Proportions of each bacterium in continuous-culture devices. P values were calculated to compare experimental and control groups. *, P < 0.05; **, P < 0.01; ***, P < 0.001. Error bars, standard deviations. Concentrations indicate levels of tetracycline.

DISCUSSION

Bacterial infection had long been one of the top causes for human death prior to the discovery of antibiotics, one of the key moments in modern science and medicine. Since then, antibiotics have been one of the most important medicines in all medical sectors. This is the reason why the development of AMR in bacteria has been such a grave concern worldwide (2). It is worth noting that AMR can be easily disseminated, which can lead to a much faster development of AMR in pathogens and can potentially lead to the enrichment of AMR, as demonstrated in previous reports where multiple MDR plasmids are present in one clinical pathogenic bacterium that develops a extensive drug resistance phenotype (60). Therefore, we believe that the dissemination of AMR is even more dangerous than the evolution of new resistance mechanisms, and this is why we are attempting to analyze dissemination dynamics of AMR in this work. This belief is further confirmed by the finding in this work that invasion of MDR E. coli, rather than the extended use of antibiotics, is primarily responsible for the development of the tetracycline resistance phenotype for S. enterica.

The dissemination of AMR takes place in a variety of ecosystems, which include wastewater treatment plants, surface waters, medical facilities, and human microbiomes such as the intestinal tract. Due to the imminent threat of AMR to human health, AMR dissemination that is taking place inside humans is the most detrimental, particularly in scenarios where AMR disseminates to bacterial infections that are being treated by antibiotics. The three AMR dissemination systems in this work were designed to simulate this scenario by assaying how AMR disseminates from an external invading MDR bacterium (E. coli) into a pathogen (S. enterica) during antibiotic treatment. Continuous-culture devices were used for this simulation, because they are great in mimicking systems where continuous input and output are constantly taking place, such as the intestinal tract. Using simulated gastrointestinal tract systems would of course be more realistic for this scenario, but that would almost certainly complicate the dynamics and may cover up mechanisms when delicate analysis is going on. The use of continuous-culture devices in this work is therefore based on considerations of both the power of mimicking real dynamics and the simplicity for analysis in order to find patterns behind phenomena. The three systems in this work represent an MDR E. coli-S. enterica cocolonization event, a post-antibiotic-treatment MDR E. coli invasion event, and an MDR E. coli invasion event prior to antibiotic treatment. Such an effort for probing AMR dissemination dynamics using continuous-culture devices has not been reported before.

The key question in the analysis of AMR dissemination is what drives the dissemination of AMR under antibiotic pressure. Both VGT and HGT are involved in this dissemination. If the acquirement of AMR leads to better competitivity under antibiotic pressure, the proportion of antimicrobial-resistant bacteria increases, leading to increased dissemination of AMR via VGT. On the other hand, antibiotics may promote transfer of ARGs between cells via HGT, leading to AMR dissemination. Transmission via VGT differs from that via HGT in that VGT directly leads to selection, which further results in a characteristic increase of one group of bacteria at the cost of the other groups. During the MDR E. coli-S. enterica cocolonization event and the MDR E. coli invasion event prior to antibiotic treatment, increased concentrations of antibiotics used in treatment led to increased MDR S. enterica contents, increased MDR E. coli contents, and decreased sensitive S. enterica contents, which strongly suggests that VGT-driven selection is the primary driver for AMR dissemination. On the contrary, during the post-antibiotic-treatment MDR E. coli invasion event, HGT apparently plays the major role here, as all groups increased following E. coli invasion, and the rate of increase for MDR S. enterica followed the same pattern as for conjugation frequencies under different antibiotic concentrations. Therefore, both VGT and HGT drive AMR dissemination but under different scenarios. During the post-antibiotic-treatment MDR E. coli invasion event where HGT is dominant, E. coli invasion took place when the total number of cells in the community was approximately 10⁵ to 10⁶ CFU/ml. During the MDR E. coli-S. enterica cocolonization event and the MDR E. coli invasion event prior to antibiotic treatment, E. coli invasion took place when the total number of cells in the community was approximately 10⁷ to 10⁹ CFU/ml. Collectively, these findings are in agreement with the proposal that HGT drives AMR dissemination when resources are not limited, while VGT drives AMR dissemination when resources are limited.

One particularly interesting finding in this work is that even environmental levels of tetracycline can lead to AMR dissemination in an HGT-driven event at a rate higher than with high therapeutic levels of tetracycline (Fig. 5 and 7). Environmental antibiotic contamination has been a long-standing and universal problem in the past few decades due to the aggressive use of antibiotics in both medical and agricultural sectors (61–63). Contamination of antibiotics has been found in a variety of environments that include human feces (64), food (65), rivers (56, 58, 59, 66), and drinking water (67). Concerns have been raised that the contamination of antibiotics may lead to the rise of AMR, but the levels of antibiotics in the environment are generally too low to elicit meaningful stress on bacteria, therefore making evolution of antibiotic resistance difficult. Findings of this work show that although the above-mentioned statement may be true, environmental levels of tetracycline can indeed lead to an increased rate of HGT, thereby promoting AMR dissemination. Therefore, the detrimental impact of environmental antibiotic pollution, particularly where direct human consumption is possible (such as in drinking water, raw consumed food, etc.), is experimentally confirmed in this work.

Both MDR and tetracycline-sensitive S. enterica strains significantly increased under therapeutic concentrations of tetracycline following MDR E. coli invasion after antibiotic treatment (Fig. 5D and E). This phenomenon is somewhat confusing, because resources in the system decreased as MDR S. enterica started to emerge, and sensitive S. enterica received the same stress as before MDR E. coli invasion. We suspect additional unidentified mechanisms play a role here in increasing the resistance levels of sensitive S. enterica to a point where they do not meet the microbiological criteria of “tetracycline resistant” but can tolerate tetracycline to a higher extent. The bacterial chromosome encodes efflux pumps, such as the AcrAB-TolC efflux pump, that may function in pumping antibiotics out of the cells (68). In the presence of antibiotics, these pumps may be upregulated, helping sensitive bacteria increase their tolerance to antibiotics, although not to the point of being resistant. Further investigations are still needed to confirm these speculations.

Antibiotic therapy is by far the best and sometimes the only option for bacterial infection treatment. This is why MDR pathogens can be such trouble when the effectiveness of antibiotics is severely hampered. In this work, by comparing AMR dissemination and pathogen growth dynamics, we found that if MDR E. coli invasion takes place prior to antibiotic treatment, the effectiveness of antibiotic treatment is reduced to near zero (Fig. 8). It needs to be noted that the MDR E. coli used here is largely considered commensal rather than pathogenic, and relatively little attention has been paid to it under clinical settings. However, when we consider a scenario where nosocomial spread of MDR E. coli takes place, although E. coli does not directly affect the health of the infected patient, it may completely sabotage the antibiotic therapy for the infected pathogen. Therefore, early invasion of MDR bacteria needs to be prevented to retain the effectiveness of antibiotic therapy, and antibiotics need to be used as early as possible in a hospital setting to avoid early nosocomial invasion of MDR bacteria. This finding here stresses the danger of MDR nonpathogens in clinical settings, where a high prevalence of these bacteria is expected in hospitals.

In conclusion, continuous-culture devices were applied for the simulation of different MDR E. coli invasion events to S. enterica communities in order to assess the dissemination dynamics of AMR. Various levels of antibiotic pressure were applied to observe the impact on AMR dissemination. Several conclusions can be made from observations obtained from this work: both VGT and HGT play roles in promoting AMR dissemination, although the level of impact differs under different scenarios, environmental contamination levels of tetracycline can drive HGT of AMR and promote dissemination of AMR to a level high enough for a pronounced effect and therefore should be given enough attention, and early MDR bacterial invasion prior to application of antibiotics may completely sabotage the effectiveness of antibiotic therapy, while MDR bacterial invasion after antibiotic application only has a partial impact. This report introduced the use of continuous-culture devices for the investigations of AMR dissemination dynamics, showed the different roles of VGT and HGT in this process, found evidence of the detrimental impact of trace-level antibiotic contamination, and raised suggestions on the early use of antibiotics in order to avoid nosocomial AMR dissemination and promote the effectiveness of antibiotic therapy.

MATERIALS AND METHODS

Strains, antibiotics, and medium.

S. enterica H9812 and E. coli BW25113 are laboratory-maintained strains. The donor strain E. coli C600(pRP4) was kindly donated by Jianqun Lin of Shandong University (69). It contains the transferable pRP4 plasmid, which carries genes encoding resistance to tetracycline, ampicillin, and kanamycin. Transconjugant S. enterica H9812(pRP4) was selected from salmonella-shigella (SS) agar medium (the final concentrations of spectinomycin, ampicillin, tetracycline, and kanamycin in the medium were 100, 100, 30, and 50 mg/liter, respectively). Ciprofloxacin, erythromycin, cephalexin, trimethoprim, and chloramphenicol were also used in this study. All antibiotics were dissolved in distilled deionized water (ddH₂O).

Phylogenetic analysis.

The evolutionary relationships of donor E. coli C600(pRP4) and recipient S. enterica H9812 were determined using the neighbor-joining method with the MEGA 7.0.21 software (70). The 16S rRNA gene was amplified and sequenced using primers shown in Table 1.

TABLE 1.

Primers for qPCR or PCRs in this study

Category	Target gene	Primer name	Primer sequence (5′→3′)	Product size (bp)	Tma (oC)	Reference
Tetracycline resistance genes	tetA	TetA-F	TCTGGTTCACTCGAACGACG	431	55	This study
		TetA-R	AGCCCGTCAGGAAATTGAGG
	tetR	TetR-F	CTCGGTCCTTCAACGTTCCT	460	55	This study
		TetR-R	TGCTGGCGGAGAATCATACG
Ampicillin resistance gene	blaTEM	TEM-F	ATTTCCGTGTCGCCCTTAT	759	52	71
		TEM-R	CTACGATACGGGAGGGCTTA
		qTEM-F	GCKGCCAACTTACTTCTGACAACG	247	60	72
		qTEM-R	CTTTATCCGCCTCCATCCAGTCTA
16S rRNA gene	16S rRNA	K90	GAGAGTTTGATCCTGGCTCAG	1,400	56	73
		K94	CGGCTACCTTGTTACGACTTC
		q16S-F	CCCAGATGGGATTAGCTTGT	106	60	74
		q16S-R	TCTGGACCGTGTCTCAGTTC

^aT_m, melting temperature.

Conjugation assays and MIC determination.

To confirm the conjugative ability for the transfer of the pRP4 plasmid from E. coli C600 to S. enterica H9812, conjugation assays were performed as previously described (75). Successful transconjugants were screened on SS agar plates supplemented with 100 μg/ml ampicillin, 30 μg/ml tetracycline, 50 μg/ml kanamycin, and 100 μg/ml spectinomycin. For determination of MICs, three colonies of transconjugants were randomly picked from each trisection of the plates and incubated in 3 ml of LB medium at 37°C for 16 h. MICs of donor, recipient, and the selected transconjugants were determined against antibiotics using the agar dilution method as previously described (76), and E. coli ATCC 25922 was used as the reference strain according to the CLSI standard.

CFU determination.

In this study, we adjusted the drop plate method for CFU determination according to the previously described method (77). After gradient dilution of fresh bacterial cultures using 0.1 M phosphate-buffered saline (pH 7.2), a 10 μl dilution was vertically dropped onto sterile plates prepared 6 h in advance, and three to six replicates were performed for each concentration gradient. The plates were then incubated at 37°C for 16 h prior to CFU counting.

Determination of conjugation efficiency.

The conjugation efficiency was determined according to a previous published method (24). Overnight cultures were resuspended in M9 minimal medium. Donor and recipient cells were mixed in a 1:1 ratio to a final volume of 800 μl and incubated at room temperature (25°C) for 1 h without shaking. The numbers of the donor, recipient, and transconjugant cells were calculated by plating on LB agar plates containing tetracycline, kanamycin, and ampicillin, SS agar plates containing spectinomycin, and SS agar plates containing tetracycline, kanamycin, ampicillin, and spectinomycin. Four to six replicate measurements were performed. Plates were incubated overnight at 37°C, and CFU were counted the following day. All experiments were repeated in triplicates. According to a previous study (24), we quantified the combined levels (in CFU) of E. coli and S. enterica before and after a 1-h incubation in the absence of antibiotic to measure the extent of growth during the mating period and the level of recipient S. enterica before and after a 1-h incubation in the presence of antibiotic to measure the extent of death. The conjugation efficiency (η_c) was calculated as follows:

$${\eta }_{\text{c}}=\frac{\text{transconjugant}}{\text{donor}\times \text{recipient}\times \text{t}}=\frac{\text{transconjugant\hspace{0.17em}}(\frac{\text{cells}}{\text{ml}})}{\text{recipient\hspace{0.17em}}(\frac{\text{cells}}{\text{ml}})\times \text{\hspace{0.17em}donor\hspace{0.17em}}(\frac{\text{cells}}{\text{ml}})\times 1(\text{h})}=\frac{\text{transconjugant}}{\text{donor}\times \text{recipient\hspace{0.17em}}}{\left(\text{cells}/\text{ml}\right)}^{-\text{1}}{\text{h}}^{-\text{1}}.$$

Continuous-culture device assembly.

The home-made continuous-culture device (Fig. 1) was assembled with the following parts: thermostats (Shenzhen Weierhai Electronics Co., Ltd., China), multiple electromagnetic stirrers (Jiangsu Jinyi Instrument technology Co., Ltd., China), and microflow peristaltic pumps (Baoding Zhunze Precision Pump Manufacturing Co., Ltd., China). Four 20-liter glass bottles each containing 12 liters sterilized fresh supplementary LB medium were prepared as medium storage. Twelve single-stage glass fermentation vessels with a total volume of 250 ml and a working volume of 60 ml were used for fermentation. Each 20-liter glass bottle provides fresh culture medium for three parallel glass fermentation vessels simultaneously. All bottles were fully covered by aluminum foil to avoid light-mediated degradation of tetracycline. The fresh LB media were pumped into the fermentation vessels at a constant rate (6 ml/h), and overflow medium at a constant rate (6 ml/h) passed from the vessels into a collecting device. The temperatures (37 ± 1°C), pH values (7.0 ± 0.5), and speeds (250 ± 20 rpm) were measured and controlled. After running the continuous-culture device for 4 h, the cultures of S. enterica H9812 or the mixtures of E. coli C600(pRP4) and S. enterica H9812 (at a cell count ratio of 1:10⁵) were inoculated into 12 vessels. Exposure to tetracycline at corresponding concentrations was carried out at the appropriate time. Sampling was performed every 12 h from 12 vessels. The running time of the continuous-culture devices was 100 to 240 h. The levels of total bacteria were measured by plating on LB agar plates. The levels of S. enterica were measured by plating on SS agar plates containing spectinomycin. The levels of MDR S. enterica were measured by plating on SS agar plates containing spectinomycin, tetracycline, kanamycin, and ampicillin. The levels of tetracycline-sensitive S. enterica were calculated by deducting MDR S. enterica levels from total S. enterica levels. The levels of MDR E. coli were measured by plating on MacConkey agar plates containing tetracycline, kanamycin, and ampicillin. Since MacConkey agar plates also select for S. enterica strains, the CFU levels of MDR S. enterica were deducted from MDR E. coli levels.

Analysis of relative content of pRP4 plasmid.

Real-time PCR (qPCR) assays was used for the determination of relative content of pRP4 plasmid according to previously published protocols (78). Primers used for PCR and qPCR are shown in Table 1. qPCR analyses of the 16S rRNA gene and bla_TEM gene were performed using a SYBR green-based approach on a StepOne Plus real-time PCR system (Applied Biosystems, USA). The levels of bla_TEM and 16S rRNA genes were quantified according to the standard curves (see Fig. S4 in the supplemental material). The relative content of bla_TEM represents the relative content of pRP4 plasmid. Three biological and three technical replicates (a total of 9 replicates) were performed for each sample.

Statistical analysis.

Independent sample two-tailed t tests were used to assess the significance of results. P values of less than 0.05 were considered statistically significant.