Central Veterinary Institute, part of Wageningen UR, P.O. Box 65, 8200, AB Lelystad, the Netherlands

Department of Farm Animal Health, Faculty of Veterinary Medicine, Utrecht University, P.O. Box 80163, 3508, TD Utrecht, the Netherlands

Department of Infectious Diseases and Immunology, Faculty of Veterinary Medicine, P.O. Box 80165, 3508, TD Utrecht, the Netherlands

Abstract

Background

Commensal bacteria are a reservoir for antimicrobial-resistance genes. In the Netherlands, bacteria producing Extended Spectrum Beta-Lactamases (ESBL) are found on chicken-meat and in the gut of broilers at a high prevalence and the predominant ESBL-gene is the _{CTX-M-1} located on IncI1 plasmids. We aim to determine the fitness costs of this plasmid for the bacterium.

We investigated the conjugation dynamics of IncI1 plasmids carrying the _{CTX-M-1} gene in a batch culture and its impact on the population dynamics of three

Parameters were estimated from experiments with pure culture of donors, recipients and transconjugants and with mixed culture of donors and recipients with a duration of 24 or 48 hours. Extrapolation of the results was compared to a 3-months experiment in which a mixed culture of recipient and transconjugant was regularly diluted in new medium.

Results

No differences in estimated growth parameters (^{4} times larger than that of the donor. In the 3-months experiment, the proportion of transconjugants did not decrease, indicating no or very small fitness costs.

Conclusions

_{CTX-M-1} gene imposes no or negligible fitness costs on its

Background

Due to the resistance against a wide range of antimicrobials including important ones such as penicillins and all cephalosporins

In Enterobacteriaceae ESBL-genes are mostly plasmid mediated and may be located on various plasmid types. In Dutch poultry _{CTX-M-1} is the predominant ESBL-gene, located on IncI1 plasmids

Although in general a high selective pressure by use of antimicrobials exists in broiler chickens, the reservoir role is unexpected in this particular case. Mass treatment of broiler chickens with cephalosporins is forbidden in the Netherlands. Cephalosporins are, however, used in one-day old reproduction animals in the poultry sector

The IncI1 plasmid is conjugative, and conjugation could explain the high abundance of bacteria carrying this plasmid in the microbiota of broilers. Within the microbiota, plasmids might act as infectious agents, which are able to persist by transfer to new bacterial hosts. Maintenance of a population of plasmids is determined by the balance between increase of bacteria carrying plasmids due to conjugation and a decrease by loss of the plasmid from bacteria and selective disadvantage of bacteria by carrying a plasmid

We aim to determine the fitness costs of this plasmid for the bacterium. Here, we used _{CTX-M-1} can persist

First, we estimated the bacterial growth parameters, conjugation coefficients and plasmid loss rate from experiments with a short duration (i.e. 24 or 48 hours). Then, we compared single and mixed cultures to determine selective disadvantage and a difference in conjugation coefficients between the donor and the newly acquired transconjugant strain

Methods

Bacterial isolates and plasmids

All isolates used in the _{CTX-M-1} on an IncI1 plasmid of sequence type 7, and is therefore resistant to cefotaxime. Isolate E75.01 was used as recipient (_{CTX-M-1} from E38.27, and is resistant to ciprofloxacin due to the presence of mutations in the chromosome (present in strain E75.01) and to cefotaxime due to the presence of _{CTX-M-1} on the obtained incI1 plasmid. Before use, transconjugants were kept in buffered pepton containing 30% glycerol at -80°C. The donor E38.27 contained a second plasmid IncHI1, which was not transferred to the transconjugant T38.27. Resistance phenotypes of

**Isolates: Characteristics of broiler ****
E. coli
**

Click here for file

The IncI1 plasmid of E38.27 contains two addiction factors

Experimental set up

Three experiments were carried out. Firstly

**Experiments: Strains and initial concentration in the experiments.** Descriptive table of the experiments in this study. Listed are the strains and initial concentrations for each experiment and the parameters estimated from these experiments.

Click here for file

In experiment 1 growth curves of single populations of ^{2} and 10^{6} cfu/ml made in 25 ml Luria Bertani (LB) broth. Start concentrations were determined directly at the start of incubation by a colony count. The flasks were incubated at 37°C. Enumerations of ^{a,b,c,d}), ^{e,f,g}) and ^{h,i,j}) were done by serial dilutions on selective plates. For the experiments with start concentration 10^{2} cfu/ml this was done at 0, 2, 4, 6, 8, 24, 30 and 48 h after the start of the experiment, whereas for the experiments with start concentration 10^{6} cfu/ml at 0, 1, 2, 3, 4, 6, 8, 24, 30 and 48 h after the start of the experiment. The growth rate, maximum density and lag-phase parameters were estimated from these data as described below in the section on the parameter estimation.

Plasmid loss was determined along with the growth experiment of ^{i}). At 4, 8 and 24 h, 94 colonies taken from the colony count plates of

Two experiments were carried out with mixed populations of ^{8} cfu/ml suspension of ^{8} cfu/ml suspension of ^{a} samples were taken for colony counts by serial dilution at 0, 3, 6, 16, 19 and 24 h after the start of the experiment. In experiment 2^{b}, two parallel series were conducted. In the first series samples for colony counts by serial dilution were taken at 0, 2, 4, 6, 8, 24, 30 and 48 h and in the second series at 0, 16 and 24 h; because of logistic reasons these sampling times were not the same.

In experiment 3, 10^{5} cfu/ml ^{2} cfu/ml

Mathematical model

The populations of bacteria growing in isolation (

**Model details: Model equations, overview of model parameters, re-parameterization of an existing growth model and derivation of specific estimators.**

Click here for file

The model to analyze the conjugation experiments contains three bacterial populations: Donor _{
D
} for the donor-recipient conjugation and _{
T
} for the transconjugant-recipient conjugation. A simpler model was also investigated in which both conjugation coefficients were assumed to be equal (_{
D
} = _{
T
}).The conjugation coefficient is defined as the number of conjugation events per bacterium per hour.

Flow diagram of the model with plasmid donor

**Flow diagram of the model with plasmid donor ****, recipient ****and transconjugant **** T.** Parameters

Plasmid loss occurs at a probability _{
CS
}. We will refer to this model as the Constant Segregation model (CS model),and (2) the rate of cell death is zero, and the rate of cell division is density-dependent. That means that the plasmid loss occurs at a rate

Long term behaviour of this system of batch cultures which were regularly diluted, was studied by applying the conjugation model for each round of the batch culture. We excluded the presence of a donor (

Parameter estimation and model selection

All estimations were done by least-squares fitting of the data (log-scaled) to the numerically solved model equations, in Mathematica (version 9,

We estimated the parameter values of _{
D
}, _{
T
}, and

The first step of the parameter estimation process was estimation of the intrinsic growth rates ^{a-j} and separately for mixed culture experiments 2^{a-b}. The estimates of the growth parameters from experiments 2^{a-b} were used for the estimation of the conjugation coefficients (_{
D
} and _{
T
}) and in the simulation of the long term experiment (see section Long term behaviour), because these experiments were also mixed culture experiments.

We fitted the model with separate _{0} and the lag-phase parameter

The second step was estimation of the rate of plasmid loss from experiment 1^{i}. From this culture 94 colonies were selected and tested for the presence of the plasmid at 4, 8, and 24 h. The number of 94 colonies was chosen for practical reasons. To estimate the plasmid loss parameters we assumed that the rate of conjugation is negligible when the population without plasmid is very small. Furthermore based on the results of experiments 1^{a-j} (Table

**Parameter**

**Value**

**95% confidence interval**

**AICc***

*AICc = Akaike’s Information Criterion (AIC) corrected for a finite sample size

AICc

**Estimate for experiments with a start culture of 10^{2} cfu/ml.

***Estimate for experiments with a start culture of 10^{6} cfu/ml.

The full model estimates different parameters _{
0
} based on the concentration of the start culture. The best fitting model was the one with different parameters for the initial concentration and lag-phase based on the concentration of the start culture, but with equal parameters for the other parameters of

Best fitting model

-19.36

2.04

h^{-1}

(1.95 – 2.14)

9.1 10^{8}

cfu/ml

(8.0 10^{8} – 10.4 10^{8})

^{2**}

0.71

h

(0.41 – 1.08)

^{6***}

1.30

h

(0.90 – 1.72)

_{
0
} 10^{2**}

0.8 10^{2}

cfu/ml

(0.5 10^{2} – 1.2 10^{2})

_{
0
} 10^{6***}

0.9 10^{6}

cfu/ml

(0.5 10^{6} – 1.6 10^{6})

Full model

-15.13

_{
R
}

2.04

h^{-1}

(1.95 – 2.14)

_{
T
}

2.09

h^{-1}

(2.00 – 2.19)

_{
D
}

2.09

h^{-1}

(2.00 – 2.19)

_{
R
}

10.7 10^{8}

cfu/ml

(8.2 10^{8} – 58.6 10^{8})

_{
T
}

10.0 10^{8}

cfu/ml

(7.0 10^{8} – 14.3 10^{8})

_{
D
}

7.6 10^{8}

cfu/ml

(5.3 10^{8} – 10.9 10^{8})

^{2**}

0.71

h

(0.41 – 1.08)

^{6***}

1.28

h

(0.89 – 1.70)

_{
0
} 10^{2**}

0.8 10^{2}

cfu/ml

(0.5 10^{2} – 1.2 10^{2})

_{
0
} 10^{6***}

0.9 10^{6}

cfu/ml

(0.5 10^{6} – 1.6 10^{6})

The sensitivity of the estimated plasmid loss parameter _{
DS
} of the DS model for the estimates of the intrinsic growth rate

The third and final step was estimation of the conjugation coefficient from experiments 2^{a-b}.

We estimated either two separate conjugation coefficients _{
D
} and _{
T
} for the donor and for the transconjugant, or a single conjugation coefficient for both (_{
D
} = _{
T
}).

Long term behaviour

For the long term behaviour of the system, we simulated the outcomes of the population dynamics for a situation in which the populations are regularly diluted 10 000 times and transplanted to new medium. This was done for either 24 h intervals or 48 h intervals. The initial concentration of the first round was _{
0
} = 10^{5} and _{
0
} = 10^{2}. We used the parameter estimates from the mixed culture experiment 2 only, because the simulation also concerned a mix of R and T.

The results of the simulations were compared to those of the long term experiment (experiment 3). We simulated five scenarios: no fitness costs (basic model), a lower growth rate of

For the two scenarios with a lower growth rate or a lower maximum density of

**Parameter**

**Value**

**95% confidence interval**

1.86

h^{-1}

(1.49 – 2.33)

9.33 10^{8}

cfu/ml

(7.79 10^{8} – 11.2 10^{8})

1.17

h

(0.70 – 1.64)

_{
0
}

2.51 10^{6}

cfu/ml

(1.75 10^{6} – 3.60 10^{6})

Results

Parameter estimates

In Table ^{a-b} (Table

**Other fits: Fitted models.** Fit results of other model structures and parameterizations.

Click here for file

All 94 samples from experiment 1^{i} at each of the three times points (4, 8 and 24 h) contained the plasmid. For both the CS model and the DS model the estimates of the plasmid loss parameters are 0.00 with one-sided 95% upper limit for the CS model probability _{
CS
} of 0.0003 per cell division, and a one-sided 95% upper limit for the DS model probability _{
DS
} of 0.0012 per cell division.

The estimate of the upper limit for the plasmid loss probability _{
DS
} in the DS model depends on the intrinsic growth rate and maximum density. Sensitivity analysis showed that this upper limit differed between 0.0008 and 0.0036 per cell division when both the intrinsic growth rate and maximum density were either a tenfold larger or tenfold smaller.

From experiments 2^{a} and 2^{b}, conjugation coefficient _{
D
} was estimated at 2.4 10^{-14} bacterium^{-1} h^{-1} (1.0 10^{-14} – 6.0 10^{-14}) and conjugation coefficient _{
T
} was estimated at 4.4 10^{-10} bacterium^{-1} h^{-1} (3.1 10^{-10} – 6.3 10^{-10}). These estimates had a better fit to the data compared to a model with the same conjugation coefficient for donor and recipient (Table

**Parameter**

**Value**

**95% confidence interval**

**AICcc***

*AICc = Akaike’s Information Criterion corrected for a finite sample size

AICc

_{
D
} _{
T
}

36.8

2.2 10^{-13}

(6.6 10^{-14} – 7.6 10^{-13})

_{
D
} _{
T
}

23.4

_{
D
}

2.4 10^{-14} 4.4 10^{-10}

(1.0 10^{-14} – 6.0 10^{-14})

_{
T
}

(3.1 10^{-10} – 6.3 10^{-10})

Experimental data on log-scale with 95% confidence intervals from experiments 2^{a – b} with mixed cultures of donor _{,} recipient

**Experimental data on log-scale with 95% confidence intervals from experiments 2**^{a – b }**with mixed cultures of donor ****, ****recipient ****and transconjugant ****.** The best fitting model (see Table

Long term behaviour

Of the five simulation scenarios, a decline of the fraction of transconjugants was found only for the scenario with a large difference in maximum density

Observed fraction of transconjugants in the bacterial population (T/(T + R) ) from long term experiments 3^{a} and 3^{b} diluting 10,000 times every 24 h (left) or 48 h (right)

**Observed fraction of transconjugants in the bacterial population (T/(T + R) ) from long term experiments 3**^{a }**and 3**^{b }**diluting 10,000 times every 24 h (left) or 48 h (right).** The dashed black line and coinciding dashed gray line describe the prediction of the simulation model for maximum density _{T} being a fraction of 0.90 and 0.95 of the maximum density _{R} The solid gray line describes the prediction for maximum density _{T} being a fraction of 0.80 of _{R}.

Also, the experimental results of the long term experiment 3 did not show a decrease in the proportion of

Discussion

Fitness costs resulting in a lower bacterial growth rate or a lower maximum density due to the presence of the plasmid IncI1 carrying the _{CTX-M-1} gene were not observed here. No differences were found between donor ^{a-j}. Fitness costs might have arisen in a competition setting with mixed populations of ^{a-b}, we could not find a difference in growth parameters between the recipient

San Millan

The extrapolation of

Plasmid loss was not observed as expected because of the presence of two addiction systems, which account for stable inheritance of the plasmid to daughter cells

Conjugation was modelled as a mass action process, which is often used to describe the spread of infectious diseases among host individuals

For our analyses, we used a logistic growth model by Barany and Roberts

The conjugation coefficient _{
T
} of the transconjugant was found to be much higher than that of the donor. This might be due to repression of conjugation

The results of this study, although obtained _{CTX-M-1} gene does not impose or only imposes small fitness costs in the absence of antimicrobials. Apart from abandoning the use of antimicrobials, additional measures might be required to reduce the occurrence of this plasmid, such as competitive exclusion with other bacteria carrying incompatible plasmids

Conclusions

Fitness costs in the absence of antimicrobials for _{CTX-M-1} gene were not found. The plasmid persisted in an

Competing interest

The authors declare that they have no competing interests.

Authors’ contribution

EF conceived the study, performed the mathematical modelling and statistical analyses, and drafted the manuscript. AvE performed the experiments. CD participated in the design of the experiments and supported the execution of the experiments. HvR participated in the design of the study, coordinated the project and helped to draft the manuscript. AS conceived the study and participated in the design of the study. DM conceived the study, participated in the design of the experiments and coordinated the experimental work. DK conceived the study, participated in the mathematical modelling and statistical analyses, and helped to draft the manuscript. All authors read and approved the final manuscript.

Acknowledgements

This work was supported by ZonMW, The Netherlands Organisation for Health Research and Development, within the Priority Medicines ‘Antimicrobiële Resistentie’ program, project number 50-51700-98-010. We thank Dr Hilde Smith of the Central Veterinary Institute, part of Wageningen UR, for explaining the addiction systems in the IncI1 plasmid. We thank three anonymous reviewers for their useful comments on a previous version of this manuscript.