In vitro conjugation kinetics of ESBL-producing Escherichia coli donors and various Enterobacteriaceae recipients


 Background: Extended spectrum beta-lactamase (ESBL)-producing enterobacteria pose a major hazard to public health. Due to the possibility of genetic transfer, ESBL genes might spread to pathogenic enterobacterial strains. Thus, information on possible genetic transfer between enterobacteria is of high interest. It was therefore the aim of this in vitro study to screen the capacity of a wide range of Enterobacteriaceae for time dependent differences in conjugation with five ESBL-producing Escherichia (E.) coli strains. Results: Conjugation frequencies for five potential E. coli donor strains, producing the enzymes CTX-M-1, CTX-M-15, SHV-12, TEM-1, TEM-52 and CMY-2, and six potential recipient strains (E. coli, Serratia marcescens subsp. marcescens, Enterobacter cloacae, Salmonella Typhimurium and Proteus mirabilis) were obtained. Hence, different combinations of donor and recipient strains were co-incubated for between 0 and 22 hours and spread on selective agar. Conjugation frequencies were calculated as transconjugants per donor. Some of the donor and recipient strain combinations did not show plasmid transfer within 22 hours. Hence, the recipient Proteus mirabilis did not accept plasmids from any of the given donors and E. coli ESBL10716 was not able to transfer its plasmid to any recipient. Enterobacter cloacae only accepted the plasmids from the donors E. coli ESBL10708 and E. coli ESBL10716 while E. coli ESBL10708 did not transfer its plasmid to Serratia marcescens subsp. marcescens. E. coli IMT11716 on the other hand did not perform conjugation with the donor E. coli ESBL10689. The remaining mating pairs differed in conjugation frequency, ranging from log -5 to -8.5 transconjugants/donor. The earliest conjugation events were detected after 4 hours. However, some mating pairs turned positive only after 22 hours co-incubation.Conclusion: The results of this study suggest that conjugation is a frequent event in the spread of ESBL genes among commensal and pathogen bacteria. This should be considered when addressing antibiotic resistance issues.


Background
Consequential to the global increase of multidrug resistant bacteria, severe economic and public health related costs have been predicted to rise signi cantly in the near future (1). In this context, extendedspectrum ß-lactamase (ESBL)-producing Enterobacteriaceae were identi ed as one of the antibiotic resistant bacterial groups currently posing the highest threat on public health (2). These bacteria have been detected in humans and animals equally. Within livestock, the highest prevalence of ESBLproducing bacteria was observed in poultry (3). ESBL-producing enterobacteria have been detected ubiquitous in poultry droppings and meat as well as the environment surrounding poultry (4,5). CTX-M-1, SHV-12 and TEM-52 are the ESBL-types most commonly detected in European chicken with Escherichia coli and Salmonella spp. as the most common bacterial hosts. Often, these enterobacteria belong to commensal bacterial populations in animals. As ESBL-producing enterobacteria are most often nonpathogenic, no clinical signs or impact on the performance are observed (3).
ESBL encoding genes are generally located on plasmids, which can be transferred between bacterial strains and species, including pathogenic strains (4,6). Thereby, harmless, unnoticed colonization with ESBL-producing bacteria can lead to diseases, which are hard to cure with antibiotics, if the recipient happens to also carry pathogenic traits. It is known for some ESBL-carrying plasmids that they do not cause tness costs to the bacterial host and can thus be passed on for generations, even in the absence of antibiotics (7)(8)(9). A transfer of ESBL-carrying bacteria from animal to human with a shared reservoir has been suggested (10)(11)(12). Antibiotic resistant bacteria may spread to humans by direct or indirect contact with animals, their food product, fecal matter or manure (4,13,14). The introduction of a TEM-52carrying E. coli from poultry to the microbial community of a human stool sample resulted in the establishment of the strain as well as plasmid transfer to an E. coli of human origin. Both donor and transconjugants were present at a lower concentration than the human bacterial strains. A simulated treatment with a selective antibiotic substance (cefotaxime) shifted the balance to the bene t of the resistant strains, which remained at high concentrations, equal to the indigenous microbiota, even days after the termination of the treatment (15). This highlights a potential pathway for resistant bacteria from animal origin to persistently colonize the human gastrointestinal tract.
This study was undertaken to investigate the time dependence of conjugation events as a rst step to estimate the possible transfer rates of ESBL genes in the intestinal tract and in the environment.
Conjugation kinetics of ESBL-carrying Enterobacteriaceae strains commonly detected in poultry were obtained in vitro within a 22-hour timeframe.

Strains and cultivation
A selection of Enterobacteriaceae strains were screened for conjugation frequencies of ESBL-carrying plasmids with ESBL-producing E. coli strains as potential donors ( Table 1). The respective ESBL-types had previously been identi ed in another project (16) and comprised CTX-M-1, CTX-M-15, SHV-12, TEM-1 and TEM-52 as well as the ampC ß-lactamase CMY-2 ( Table 2). All donor strains were isolated from broilers samples. A total of 35 enterobacterial strains (Table 1) were screened as potential recipients.
These species are commonly detected in the gastrointestinal tract of poultry (17).
The bacterial strains were stored in cryo stocks and cultivated aerobically overnight at 37 °C in Müller-Hinton broth (MHB) (Carl Roth GmbH + Co. KG, Germany) with or without antibiotic supplementation. MacConkey agar (Carl Roth GmbH + Co. KG, Germany) was used for all plates, except for an agar disc diffusion assay, where Müller Hinton agar (Carl Roth GmbH + Co. KG, Germany) was used, and incubated aerobically overnight at 37 C.

Antibiotic resistance screening
To perform conjugation trials, potential donors and recipients were chosen based on miss matching antibiotic resistance pro les. Hence, resistance and sensitivity to 20 different antibiotic substances were determined for the potential recipients (n=35) and the potential donors (n=5) by agar disc diffusion tests.
To qualify as recipients, strains had to show an inhibition zone around the 5 µg/mL cefotaxime disc (CTX) and be resistant to an additional antibiotic substance, which inhibited the growth of at least one potential donor. This allowed the detection of transconjugants on double antibiotic MacConkey agar. Recipient and donor strains for the conjugation screening were chosen by non-overlapping antibiotic resistances.
Speci cation of antibiotic resistance and susceptibility Suitable antibiotic dosages for the inhibition of the strains were determined by examination of growth kinetics in broth microdilution tests for 24 hours at 37°C.
In short, strains were pre-cultured from cryo stocks overnight in MHB without antibiotics and subsequently washed twice in phosphate-buffered saline (PBS, Sigma-Aldrich, Chemie GmbH, Germany). The cells were re-suspended in MHB without antibiotics and diluted to 105 cells/mL. Minimal inhibitory concentrations (MIC) were obtained for the relevant antibiotics by broth microdilution in duplicates. Turbidity was measured at 690 nm every 10 minutes for 24 hours in a microtiter plate reader (Name, Tecan Austria GmbH, Austria). Non-inoculated media served as negative controls, while inoculated MHB without antibiotics provided the positive control. The MIC was de ned as the concentration, at which no growth was observed within the 24-hours (h) period of measurement. According to the MIC and growth curves, antibiotic concentrations for agar plates for the conjugation trials were chosen.

Screening for conjugation
Potential donor and recipient strains were co-cultivated for 22 h to identify positive conjugation pairs (supplementary data). The screening was implemented in duplicates. To obtain viable and antibiotic resistant bacterial cells in their log-phase, bacteria strains were pre-cultured twice in MHB with antibiotics (same antibiotic concentration as in agar plates), which could not inhibit the growth of the respective strain and once in MHB without antibiotics. Thereafter they were washed twice in PBS and diluted in MHB without antibiotics to 106 cells/mL. Equal volumes (100 µL) of the donor and recipient suspensions, were added to 800 µL MHB without antibiotics. Single strain donor and recipient dilutions served as control. All samples were incubated for 22 hours and subsequently spread on double antibiotic agar plates at different dilutions. The antibiotic combinations for the different donor and recipient strains are shown in the supplementary data. Positive and negative controls were obtained by spreading the control suspensions on double (negative) and single (positive) antibiotic agar plates. All plates were incubated overnight, and conjugation events were identi ed as colony growth on double antibiotic agar plates.
Colony forming units (cfu)/mL were obtained to estimate useful dilution levels for the 22-h kinetic assay.

Kinetic assay
The conjugation events identi ed during the screening were further investigated in 22 h kinetics (supplementary data). Duplicates of the donor and recipient strains (Table 3) were precultured (MHB with antibiotics), washed and diluted to 106 cells/mL as described above. One mL of donor and recipient suspensions, respectively were inoculated in 8 mL MHB without antibiotics, mixed thoroughly, dispensed in 1.4 mL aliquots and incubated for 0, 2, 4, 6, 8 and 22 h, respectively. Inocula (1 mL, 105 cells/mL) with only one bacterial strain (donor or recipient) served as controls. Immediately after the inoculation, 300 µL of the suspension were plated on two double antibiotic agar plates to identify transconjugants present at hour 0. Simultaneously, dilution series were spread on MacConkey agar plates without antibiotics to obtain the total cell count. This procedure was repeated after 2, 4, 6, 8 and 22 h with suitable dilutions. The single-strain suspensions were plated on the corresponding double and single antibiotic agars for negative and positive controls respectively. The plates were incubated overnight and the conjugation frequency (CF) was calculated as transconjugants/donor (18).

Results
Antibiotic resistance screening of recipient strains The results obtained from the agar disc diffusion tests (supplementary data) identi ed six potential recipients, 5 potential donors and 24 donor-recipient combinations for the screening assay

Screening for suitable mating pairs
The results from the screening for conjugation are presented in Table 2. Proteus mirabilis DSM 4479, who did not show conjugation with any of the given donors, was excluded from further trials. The same applied to the potential donor E. coli 10716, who did not transfer plasmids to three potential recipients Serratia marcescens subsp. marcescens DSM 5570, Enterobacter cloacae DSM 30060 and Proteus mirabilis DSM 4479. As the remaining four donors and ve recipients proved the ability to produce transconjugants, they were further studied for the kinetic study.

Kinetic assay
Varying time depending conjugation frequencies were observed for different donor and recipient pairs within the 22 h incubation period ( In summary, observed conjugation frequencies were within the range of 10-9 -10-5 transconjugants/donor. The highest conjugation frequency was observed for the Salmonella Typhimurium recipient strain and a CTX-M-1 carrying plasmid. For the majority of the investigated strains, no transconjugants were observed after 8 h. Four of the mating pairs led to transconjugants above detection level within 4 h of co-cultivation. Differences in conjugation frequency and incubation time were observed depending on bacteria genera, species and strain.

Discussion
This in vitro study investigated the conjugation kinetics between ESBL-producing E. coli donors and various Enterobacteriaceae recipients. In this study we used relatively high concentrations of cefotaxime in the agar. This was due to the results from the initial resistance screening, where 8 µg ctx/mL did not affect the growth of the donors negatively in the broth microdilution assay. Surprisingly, these donors were inhibited by 30 µg CTX discs in the agar diffusion trial. The Clinical Laboratory Standards Institute (CLSI) suggests this disc type for screening for ESBL-producing bacteria and 1 µg ctx/mL for broth microdilution (19). Hence, the donors would have failed to be identi ed in the recommended agar disc diffusion test but easily be recognized as ESBL-producers in the broth microdilution test.
Of the possible six recipient and 5 donor strains, only six suitable mating pairs were found that could be used in kinetic mating experiments. Successful conjugation was observed in 92 -100 % of the donor/recipient combinations (51 donors, 1 recipient and 48 donors and 1 recipient) investigated (20,21) compared to 52.2 % in this present study (5 donors, 6 recipients). This rate is affected by chance and a higher number of recipient strains may have altered the outcome. Also, this study used different bacterial species while only E. coli donors and recipients were used in the study by Franiczek and Krzyzanowska (21). The results of this study suggest that the propensity of ESBL-producing donors for gene transfer differ signi cantly between strains from the same species. E. coli ESBL10682 and E. coli ESBL10717 transferred their plasmids to 4 of the possible 6 and 5 potential recipients respectively. On the contrary, E. coli ESBL10716 did not transfer its plasmid to any of the recipients provided. This suggests that the bla genes were located on the chromosome or a non-conjugative plasmid (20) in this donor. E. coli ESBL10689 and E. coli ESBL10717 were able to perform conjugation with 2 out of 4 or 3 potential recipients, respectively. A plausible reason for the cases were the recipient did not accept the plasmid is that the recipients may already harbor plasmids with the same replicon (22). Also, the incubation time might have been too short, or the initial concentrations of donor and recipients were too low (23). The latter is rather unlikely, as the initial concentration of 105 cfu/mL is quite high and the long incubation time of 22 hours in media increases cell concentrations even further. Furthermore, the transconjugants could have been under detection limit. Here, the detection limit was 3 cfu transconjugants/mL, which makes this option rather unlikely at the given bacterial concentrations and incubation time. Finally, the recipient may have speci c endonucleases which destroy the plasmids after uptake and thereby prevent the formation of transconjugants (24,25). This may be the case for the Proteus mirabilis recipient, which did not mate with any of the given donors.
When co-cultivated, the recipient strains S. marcescens and Salmonella Typhimurium showed lower growth rates than the donor strains (supplementary data). Hence, the donor/recipient ratio and subsequent conjugation frequencies was shifted towards the donor. This effect should be considered when evaluating conjugation events as conjugation frequencies per recipient cfu would have been higher than conjugation rates per donor cfu. Thus, calculation of conjugation events per donor may be biased and other methods of calculation may give different results (26, 27). However, as the present study was designed to nd model strains to study conjugation kinetics in detail, this was not the focus of the research. Hence, the calculation of transconjugants/donor was su cient to compare the different mating pairs.
Conjugation frequencies differed between various donor and recipient strains in the employed in vitro assay. Genera and strain depending variations in conjugation frequency have been described previously (18,20). In some studies, it was suggested that conjugation occurs more regularly with donor and recipients from different genera. Donor strains belonging to Enterobacter cloacae, E. sakazakii, E. agglomerans, Serratia marcescens, Citrobacter freundii and E. coli showed conjugation frequencies between 10-7 and 10-1 transconjugants per donor with the recipient E. coli K12C600 (20,21). Correspondingly, conjugation rates were rather low in the present study, when E. coli strains served as recipients (10-9-10-7 transconjugants/donor). On the contrary, Yamaichi et. al., (9) described higher conjugation frequencies for mating pairs of the same species than interspecies donor/recipient combinations. This corresponds with our ndings for the E. coli ESBL10689 donor. Thus, this study cannot con rm a general statement to either direction, but rather suggests strain speci c differences. These relatively high conjugation rates reported in the literature compared to the frequencies shown in the present study may depend on different strains used. While some studies used different strains as donors and a consistent E. coli recipient (20,21), the present study used varying donor and recipient strains. In the mentioned studies, especially Citrobacter freundii, a strain not investigated in the present study, showed high conjugation rates, while S. marcescens donors showed rather low conjugation frequencies, comparable to the results of this study.  (17,29,30). Therefore, the detection limit must be considered when evaluating the time for the rst observed conjugation event. The impact of initial concentrations of the mating pair on the number of transconjugants after a given time of coincubation and thereby the detection limit of conjugation was previously described in a study by Handel et al. (23).
The time until detection of transconjugants differed signi cantly between donors with the same recipients. Conjugation kinetics for ESBL-carrying plasmids have previously been studied, but mainly with longer time intervals (15). It was also shown in the present study that both time as well as number of The aim of this study was to identify mating pairs tted for future in vivo studies in poultry. These mating pairs should comprise a donor producing an ESBL type with high prevalence in broilers (3) and perform conjugation at bacterial concentrations commonly observed in the hindgut (17,30 Availability of data and material The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interest.
Funding disclosure E.S. received the Elsa Neumann scholarship the research work presented in this study. The Study was conducted within the EsRAM project, funded by the German Federal Ministry of Food and Agriculture.
Open Access Funding was provided by the Freie Universität Berlin.
Authors contributions W.V: and E.S. planned the experiment, which was carried out by E.S. who also wrote the manuscript with input from all authors. J.Z. helped supervise the project. W.V. and J.Z. conceived the original idea. W.V. supervised the project. All authors discussed the results and commented on the manuscript