Bacterial Species-Specific Activity of a Fluoroquinolone against Two Closely Related Pasteurellaceae with Similar MICs: Differential In Vitro Inoculum Effects and In Vivo Efficacies

We investigated the antimicrobial activity of a fluoroquinolone against two genetically close bacterial species belonging to the Pasteurellaceae family. Time-kill experiments were used to measure the in vitro activity of marbofloxacin against two strains of Mannheimia haemolytica and Pasteurella multocida with similar MICs. We observed that marbofloxacin was equally potent against 105 CFU/mL inocula M. haemolytica and P. multocida. However, an inoculum effect was observed with P. multocida, meaning that marbofloxacin activity was decreased against a 108 CFU/mL inoculum, whereas no inoculum effect was observed with M. haemolytica. Marbofloxacin activity was also tested in a lung infection model with immunocompromised mice intratracheally infected with 109 CFU of each bacteria. At the same dose, the clinical and bacteriological outcomes were much better for mice infected with M. haemolytica than for those infected with P. multocida. Moreover, bacteriological eradication was obtained with a lower marbofloxacin dose for mice infected with M. haemolytica. Our results suggest that the differential in vivo marbofloxacin efficacy observed with the two bacterial species of similar MIC could be explained by a differential inoculum effect. Consequently, MICs determined on 105 CFU inocula were not predictive of the differences in antibiotic efficacies against high bacterial inocula of closely related bacterial strains. These results could stimulate further investigations on bacterial species-specific antibiotic doses in a clinical setting.


Introduction
In veterinary and human medicine, the fight against antimicrobial resistance requires the reduction of antimicrobials consumption through new regulatory requirements, the suppression of inappropriate uses, and also through the optimization of dosage regimens for appropriate uses. Such optimization of dosing regimens should more precisely take into account factors depending on the pathogens, the infection sites, the molecules, the hosts, and their interactions [1].
Mannheimia haemolytica and Pasteurella multocida are two Gram negative bacterial species belonging to the Pasteurellaceae family, which are closely genetically related. These bacteria are the most often encountered bacterial aetiological agents of a pulmonary disease in cattle called BRD complex. This disease can be treated by several antibiotics belonging to different families, including fluoroquinolones, and for a given antibiotic the same dosage regimens are registered for the overall disease independently of the bacterial aetiological agent. This situation of "disease claim" is representative of veterinary and human treatments targeting a cluster of pathogens. In the context of prudent use of antimicrobial drugs, we wondered if such "disease claim" could potentially be improved by more deeply exploring the activity of antibiotics against the different aetiological agents.
The efficacy of an antibiotic treatment depends among diverse factors on the susceptibility of the targeted pathogens, and it is the reason why the MICs of antibiotics used against bacteria are combined with pharmacokinetic parameters to explore the efficacy of dosage regimens. For fluoroquinolones, two PK/PD indices are correlated with the effectiveness of an antimicrobial treatment: AUC/MIC and Cmax/MIC [2,3]. However, a set of studies have pointed out the limitations of MIC as the sole indicator to capture the complexity of interactions between bugs and drugs [4]. These limitations are linked to the standard procedure for its determination that consists in assessing antibiotic activity on an inoculum of 5.10 5 CFU/mL in the presence of constant concentrations of drugs and after a fixed time of incubation [5]. As an example, the standard MIC determination does not take into account the inoculum effect whereas several studies have stressed that the activity of antibiotics is decreased when they face high bacterial inocula at the infection sites [6][7][8][9][10].
By using in vitro and in vivo assays, the aim of this study was to determine the activity of marbofloxacin, a fluoroquinolone used in veterinary medicine in Europe, against two strains of two closely related bacterial species belonging to the same Familiae (Pasteurellaceae) that have similar MICs and are targeted in the field by the same dosing regimens.

Material and Methods
Bacterial strains, media, antibiotic and reagents A strain of M. haemolytica serotype A2, isolated in 2010 in the Netherlands from the lungs of a calf with clinical signs of BRD, and a strain of P. multocida isolated in a diseased pig in France were used in this study.
Mueller Hinton (MH) broth and agar and Brain Heart Infusion Broth (BHIB) were purchased from Biomerieux.
Marbofloxacin was provided by Vetoquinol SA and used for MIC and MBC determinations and for time-kill experiments. For in vitro testing, the powder was diluted in sterile distilled water from a stock solution at 1 mg/mL stored at -20°C. For in vivo testing (mice challenge), a commercial formulation of marbofloxacin for veterinary use approved in France (Marbocyl 10% 1 ) was used immediately after dilution in sterile distilled water.

In vitro testing
Determination of the Minimum Inhibitory Concentration (MIC) and the Minimum Bactericidal Concentration (MBC). MIC were determined in triplicate according to the CLSI reference methods (5). MIC were also determined in Brain Heart Infusion Broth (BHIB) with bacterial inocula of 5.10 5 CFU/mL, using the same microdilution method since BHIB gave better growth of both strains and was used for killing curves experiments.
Bactericidal effects were determined with bacterial inocula of 10 5 CFU/mL and 10 8 CFU/ mL, from time kill experiments by subculturing in MH agar plates, where wells with no visible growth were observed. The minimum bactericidal concentration (MBC) was defined as the lowest marbofloxacin concentration giving a decrease of more than 3log10 CFU/mL (99.9%) at 24h.
Time kill experiments. After overnight growth on MH agar plates, inocula of M. haemolytica and P. multocida were prepared by transferring several colonies into fresh BHIB to obtain suspensions containing 10 5 CFU/mL and 10 8 CFU/mL. Suspensions were incubated for one hour at 37°C to reach exponential growth phase before adding marbofloxacin at different concentrations (0, 0.25, 0.5, 0.75, 1, 2, 4, 8, 16 and 32 fold the MIC). Aliquots of 500 μL were removed from bacterial suspensions at 0, 0.5, 1, 2, 4, 6, 8 and 24h after incubation at 37°C. They were centrifuged at 3000g for 10 minutes and pellets were then resuspended in 500 μL of 0.9% NaCl to eliminate the carryover of marbofloxacin. Ten μL of successive 10-fold dilutions were plated on MH agar. The colonies were counted after 24h of incubation at 37°C. The detection limit of colony counting was 100 CFU/mL.
Mean Inoculum growth: From each time-kill curve, we calculated the Area Under the Curve (AUC) of bacterial counts from time 0 to 24h. This AUC was divided by 24 hours to obtain an average inoculum size over the investigated time interval (24 hours) for each x tested marbofloxacin concentration [3]. The basal inoculum (I BASAL ) was equal to the inoculum size before exposure to marbofloxacin. Mean inoculum growth (ΔI(x)) was then defined by Eq 1: where ΔI(x) is expressed in log CFU/mL, I x (log CFU/mL) is the mean inoculum obtained after exposure to a given marbofloxacin concentration x and I BASAL (log CFU/mL) is the inoculum size before exposure to marbofloxacin.
Modelling of mean inoculum growth: Mean inoculum growth was modelled as a function of marbofloxacin concentration, using a sigmoid dose-response effect model implemented in the WinNonlin software version 5.3 (Pharsight, Moutain View). We re-parameterized equations previously used elsewhere [11,12] which are derived from a classical sigmoid E max model, to obtain Eqs 2 and 3 allowing to directly calculate two parameters of interest, i.e. EC -3log the antimicrobial concentration producing a reduction of 3log 10 (99.9% reduction) of the basal inoculum of 10 5 CFU/mL, and EC -5log the antimicrobial concentration producing a reduction of 5log 10 (99.999% reduction) of the basal inoculum of 10 8 CFU/mL.
where ΔI(x) (log CFU/mL) is the mean inoculum growth for marbofloxacin concentration x, ΔI MAX and ΔI MIN (log CFU/mL) the maximal and minimal inoculum growth, obtained without antibiotic (ΔI MAX ) or after exposure to the highest marbofloxacin concentration (ΔI MIN ), x (μg/mL) the tested marbofloxacin concentration, EC -3log (μg/mL) the antimicrobial concentration producing a reduction in the basal inoculum of 3 log 10 , i.e. ΔI(EC -3log ) = -3, EC -5log (μg/ mL) the antimicrobial concentration producing a reduction in the basal inoculum of 5 log 10 , i. e. ΔI(EC -5log ) = -5, and γ the sigmoid coefficient of the curve. These values of the mean inoculum changes (-3 log 10 and -5 log 10 ) were chosen to take into account the size of the initial inoculum (5 log 10 and 8 log 10 ). In both cases, the remaining inoculum (2 or 3 log 10 ) can be considered as similar, and low enough to be eradicated in vivo by the host natural defences.

In vivo testing
Animals. All animal procedures were carried out in accordance with international accepted standards of animal care under agreement number TOXCOM/0022/AF GL (Ethic committee: Comité d'éthique de Pharmacologie-Toxicologie de Toulouse-Midi Pyrénées, France #C2EA-86) for animal experimentation of French Ministry of Agriculture.
Eighty female Swiss mice weighing 20 to 30g (Charles River Laboratories, L'Arbresle, France) and aged from 8 to 12 weeks were conditioned in our laboratory at least for one week before experimentation. Animals were housed with a 12 hours day-night cycle in a temperature (22°C) controlled room in autoclavable type II L cages with filters on Souralit 1 (hygiene animal bedding) and had water and food (Tecklad global diet 2014 Harlan 1 ) access ad libitum.
Induction of neutropenia. Cyclophosphamide (Sigma-Aldrich, St Quentin Fallavier, France) was administered intraperitoneally to mice at doses of 150 mg/kg 4 days before inoculation, and 100 mg/kg one day before inoculation.
Lung infection model. Mice were purchased specific pathogen free and the health status was checked twice a day before and during experiments. We verified mobility, respiratory movements and piloerection. If immobility was observed, we euthanized animals with the humane method described below. These animals were included in the "dead group". If piloerection and/or increase of respiratory movements were detected, the mice were more carefully observed during the day.
The tracheas were cannulated with a 22 G catheter without mandrel. The lungs were inoculated with 20μL of M. haemolytica or P. multocida suspension containing 5x10 11 CFU/mL, corresponding to a total of 1x10 9 CFU/lung. To constitute a control, 2 mice remained uninfected and were isolated in a separate box.
Three hours after bacterial challenge, mice received a single intraperitoneal injection of marbofloxacin. Eight groups of 10 to 13 mice were randomly constituted and were treated with 0, 1, 5, 10, 20 or 40 mg/kg of marbofloxacin for the mice inoculated with M. haemolytica, and with 0 or 40 mg/kg of marbofloxacin for the mice inoculated with P. multocida. For ethical reasons, only 4 to 8 mice were allocated in groups with low variability.
Clinical signs of disease and mortality were recorded over 48h, and at day 2 after challenge, all surviving mice were sacrificed humanely by an intraperitoneal injection of 120 mg/kg pentobarbital sodium (Dolethal 1 , Vétoquinol, France). The lungs were aseptically removed and homogenized in 5mL of 0.9% NaCl. The homogenates were centrifuged at 3000 g for 10 minutes and the pellets resuspended in 2 mL of 0.9% NaCl. Ten μL of successive 10-fold dilutions were then plated on MH agar. The bacterial counts per lung were recorded after 24h of incubation at 37°C. The detection limit of colony counts was 200 CFU/lung. Mice were then divided into 3 categories regarding the effect of marbofloxacin: "persistent infection" (decrease in bacterial load <3log10 CFU in lungs), "bactericidal effect" (decrease in bacterial load >3log10 CFU in lungs), "bacterial eradication" (bacterial counts below the LOQ). In the follow-up, dead mice were classified into the "persistent infection" category.
Statistical methods. The bacterial counts in lungs after 48 h were compared with an analysis of variance followed by a Dunnett comparison.  Table 1. The MIC values of marbofloxacin were 0.03 μg/mL for M. haemolytica, and 0.016 μg/mL for P. multocida. These values were identical in MHB and BHIB.

In vitro testing
The MBC values observed with the standard inoculum of 5.10 5 CFU/mL showed very slight differences, i.e. equal or less than one 2-fold dilution, compared to the corresponding MIC values for both M. haemolytica (0.023 μg/mL) and P. multocida (0.03 μg/mL), as indicated by MBC/MIC ratios (Low inoculum, Table 1). The MBC value obtained with the 5.10 8 CFU/mL inoculum was equal to the MIC (0.03 μg/mL) for M. haemolytica, but was 8-fold the MIC (0.12 μg/mL) for P. multocida (High inoculum, Table 1).
Time-kill curves. The time-kill curves of M. haemolytica and P. multocida exposed to different concentrations of marbofloxacin are presented in Fig 1. For M. haemolytica, we obtained a concentration-dependant bactericidal effect. For the low inoculum of 10 5 CFU/mL the MBC was equal to 0.75xMIC. Marbofloxacin killed 99.9% of the bacteria within 0.5h at concentrations higher than 2xMIC, within 2h at 1 xMIC and within 4h at 0.75xMIC.
For P. multocida, the results obtained with the low inoculum of 10 5 CFU/mL were similar to those observed with M. haemolytica. The MBC of marbofloxacin was equal to 1xMIC, even if the effect was slower than for M. haemolytica (killing <3log10 at 8h). In addition, marbofloxacin achieved 99.9% killing within 0.5, 1 and 2h at concentrations of 8x, 4x and 2xMIC, respectively. For the high inoculum of 10 8 CFU/mL P. multocida, the MBC was ! 8xMIC; marbofloxacin achieved 99.9% killing within 1h at concentrations ! 16xMIC, 2h at 8xMIC. At 4xMIC, a transitory bactericidal activity was observed for the first 8 hours but a regrowth was observed at 24h. Pharmacodynamic modelling. The relations between inoculum changes and marbofloxacin concentrations are presented in Fig 2. The observed data were fitted with the models described by Eqs 2 and 3 and the pharmacodynamic parameters are presented in Table 2.
For the low inoculum of both M. haemolytica and P. multocida, concentration-effect curves were close, with marbofloxacin concentrations associated with ΔI = -3 around 1xMIC (see Fig  2, upper panel). Indeed, data fitting using Eq 2 indicated that EC -3log (concentrations associated with ΔI = -3) were 0.69xMIC and 1.34xMIC for M. haemolytica and P. multocida, respectively. The ratio of EC -3log for P. multocida and M. haemolytica was 1.95 (Table 2).
For the high inoculum, the concentration-effect curve of M. haemolytica (Fig 2, lower  panel) showed that inoculum changes occurred in a concentration range of marbofloxacin similar to that observed with the low inoculum of M. haemolytica (between 0.5 and 2-fold the MIC). Conversely, the concentration-effect curve for the high inoculum of P. multocida was dramatically shifted to the right, indicating that the marbofloxacin concentrations required to achieve the same magnitude of inoculum change were much higher with P. multocida than with M. haemolytica. Data fitting using Eq 3 indicated that EC-5log (concentrations associated with ΔI = -5) were 1.03xMIC and 10.0xMIC for M. haemolytica and P. multocida, respectively ( Table 2). The ratio of EC -5log for P. multocida and M. haemolytica was 9.74 (Table 2).

Infection model
Clinical outcomes: mouse survival. Mice that were not infected showed normal activity and no death, during the whole experiment. The results obtained in the different groups of infected mice are presented in Table 3. In the two infected/non-treated groups (0 mg/kg marbofloxacin dose), all the mice infected with M. haemolytica or P. multocida presented clinical signs of disease. About 15% (2/13) of the mice infected with M. haemolytica died between 36-48 h post-challenge, and 100% (4/4) of the mice infected with P. multocida died between 12-36 h post-challenge (85% and 0% survival at 48 h, respectively, Table 3).
In the group of mice infected with P. multocida and treated with the 40 mg/kg marbofloxacin dose, 100% (11/11) of the mice showed persistent clinical signs of infection (they developed mildly scruffy coat and were prostrate) and were still alive at 48h post-challenge (Table 3).
In the groups of mice infected with M. haemolytica and treated with the 40 mg/kg marbofloxacin dose, 100% survival was observed at 48 h post-challenge and none of the mice showed any clinical signs of disease. When using decreasing doses of marbofloxacin (20, 10, 5 and 1 mg/kg) in mice infected with M. haemolytica, 100% survival was observed at 48 h post-challenge and the percentage of mice presenting clinical signs progressively increased from 20% to 100% (Table 3).
Microbiological outcomes: pulmonary bacterial loads. The lung bacterial counts measured in mice 48h after the challenge are reported in  The mean bacterial counts in the lungs of untreated animals infected with M. haemolytica and P. multocida were 8.48±0.97 and 10.08±0.06 log 10 CFU/lung, respectively.
The 40 mg/kg marbofloxacin dose led to a dramatic decrease in the mean bacterial counts for M. haemolytica infected mice (1.91±0.90 log 10 CFU/lung) compared to the untreated group (8.48±0.97 log 10 CFU/lung), whereas the mean bacterial counts were only slightly decreased for P. multocida infected mice (8.10±1.30 versus 10.08±0.06 log 10 CFU/lung). For M. haemolytica infected mice, decreasing the marbofloxacin doses from 20 to 1 mg/kg showed a progressive increase in the mean bacterial loads in the lungs at 48h post-challenge. An analysis of variance followed by a Dunnett comparison indicated that the mean bacterial count for 1 mg/ kg was not significantly different (p>0.05) from the untreated group, whereas the mean bacterial counts were significantly decreased compared to the control group for the 5 mg/kg and higher doses (p<0.05).
The 40 mg/kg marbofloxacin dose produced bacterial eradication or a bactericidal effect in 37% and 63% of mice infected with M. haemolytica, respectively. In mice infected with P. multocida, the 40 mg/kg marbofloxacin dose produced no bacterial eradication, and only a 9% bactericidal effect.
In mice infected with M. haemolytica, the decrease in marbofloxacin doses from 20 to 1 mg/ kg was correlated with an increase in the percentage of mice with persistent infection: 9%, 42%, 40% and 100% for the marbofloxacin doses of 20, 10, 5 and 1 mg/kg, respectively.

Discussion
In the present study, we assessed in a mouse model of pulmonary infection the efficacy of an antimicrobial treatment with marbofloxacin against two bacterial species belonging to the Pasteurellaceae family: M. haemolytica and P. multocida. These two bacterial species are among the etiologic agents of the same disease in cattle.
Our main result demonstrated that the dose of marbofloxacin required to cure experimental pneumonia in mice was dramatically different depending on the bacterial pathogen responsible for the infection, despite the fact that these pathogens have similar MICs and are closely genetically related (Pasteurellaceae family). Moreover, in vitro assessment of the antibacterial activity of marbofloxacin against both bacterial species demonstrated that the inoculum effect, that corresponds to the decrease of antibacterial activity when inoculum size increases, and which was previously identified as an important factor of antimicrobial efficacy of fluoroquinolone [9,13,14], was also dramatically different in magnitude between the two bacterial species.
Our first objective was to develop a rodent model of pulmonary infection with M. haemolytica or P. multocida to enable clinical and bacteriological outcomes to be assessed after marbofloxacin treatments. In our experiments, we chose to depress the immune response of the mice, by the administration of cyclophosphamide before the course of challenges [15] to get similar bacterial counts at the time of treatment with both bacteria. By minimizing the implication of the natural (specific and non-specific) defence, neutropenia enabled overt disease to be established in order to compare the effects of marbofloxacin on both bacterial species.
In this pneumonia model, we first evaluated the clinical and microbiological outcomes after marbofloxacin treatment at a dose of 40 mg/kg. This dose had already been tested in a pneumonia model with P. multocida in immunocompetent mice [16]. In the present study with immunocompromised mice infected by P. multocida, the 40 mg/kg dose prevented mortality at 48 h post-challenge, but it produced neither a clinical cure nor a significant decrease in the bacterial loads in the lungs. Conversely, the 40 mg/kg dose in immunocompromised mice challenged with M. haemolytica was associated with a 100% clinical cure rate and a significant microbiological effect in all treated mice (including a microbiological cure rate of 40%). We then decided to explore the magnitude of this differential drug efficacy by treating M. haemolytica infected mice with decreasing doses of marbofloxacin. For ethical reasons, as the 40 mg/kg dose was poorly effective against P. multocida, we did not test lower doses of marbofloxacin in P. multocida infected mice. After testing marbofloxacin doses of 20, 10, 5 and 1 mg/kg, we demonstrated that the 5 mg/kg dose was the lowest dose producing better clinical and microbiological outcomes in the M. haemolytica pneumonia than the 40 mg/kg marbofloxacin dose in the P. multocida pneumonia (Fig 3 and Table 3).
We explored in vitro using time-kill experiments the antibacterial activity of marbofloxacin against both P. multocida and M. haemolytica strains. The bacteriological results were expressed in terms of mean inoculum changes over 24 h, and analysed as a function of marbofloxacin concentrations [11,12]. With initial inocula of 10 5 CFU/mL, the marbofloxacin concentrations producing a 99.9% (-3 log 10 ) reduction in the basal inoculum were similar between the two bacterial strains and close to their respective MICs (0.69xMIC for M. haemolytica and 1.34xMIC for P. multocida, Table 2). Similar results were obtained by Ferran et al. [12] with P. multocida (2.06xMIC) and E. coli (1.80xMIC). This is in agreement with the observation that the MBC of marbofloxacin was close to the MIC for both P.multocida and M. haemolytica. Altogether, these results support the view that MIC is a relevant indicator of the marbofloxacin bactericidal effect against low bacterial inocula.
Using initial inocula of 10 8 CFU/mL, huge differences were observed when comparing the marbofloxacin activity against P. multocida and M. haemolytica strains (Fig 2, lower panel; Table 2).
For P. multocida, the concentration-response curve was shifted to the right of the x-axis, and the marbofloxacin concentration that gave a 99.999% (-5 log 10 ) reduction in the basal inoculum was about 10-fold the MIC. These results are evidence of the inoculum effect, consisting of a reduction in the antibiotic activity while bacterial inoculum size increases. This effect was previously demonstrated with P. multocida [12] and with other Gram-negative and Gram-positive bacterial species [9,13,14,17].
Unexpectedly, the concentration-response curve for M. haemolytica remained close to that obtained with the 10 5 CFU/mL inoculum, and the marbofloxacin concentrations associated with a 99.999% (-5 log 10 ) reduction in the basal inoculum was 1.03xMIC which is similar to the value obtained with a -3 log 10 reduction of the 10 5 CFU/mL inoculum (0.69xMIC). The same result was obtained with another strain of M. haemolytica (data not shown). M. haemolytica is, to our knowledge, the first reported bacterial species that does not exhibit such an inoculum effect for a fluoroquinolone.
Interestingly, the results obtained in vitro with the high inocula (10 8 CFU/mL) of P. multocida and M. haemolytica (Fig 2) mirror those obtained in vivo in the pneumonia model produced with a bacterial load of 10 9 CFU/lung. Indeed, the ratio of in vitro effective concentrations (EC -5log ) for P. multocida and M. haemolytica was equal to 9.71 (Table 2), and the ratio of in vivo doses producing similar responses for P. multocida and M. haemolytica was about 8 (40 mg/kg for P. multocida and 5 mg/kg M. haemolytica).
In a set of published works, it has been demonstrated in experimental models of infections that the antimicrobial activity of fluoroquinolones or third-generation cephalosporins could be reduced against high bacterial loads compared to low bacterial loads of different bacterial species, including Escherichia coli, Pasteurella multocida, Klebsiella pneumoniae and Staphylococcus aureus [6,9,13,18]. In all these cases, in vitro time-kill experiments have shown the occurrence of an inoculum effect.
Altogether, these data suggest that the higher activity of marbofloxacin in the pneumonia induced by a high inoculum of M. haemolytica compared to P. multocida was potentially linked to the absence of an inoculum effect in M. haemolytica. Therefore, it might be speculated that marbofloxacin was as effective against the high inoculum of M. haemolytica as it would be against a low inoculum. Unfortunately, this situation could not be tested in our model of infection because of the low pathogenicity of M. haemolytica in mice, which did not produce overt clinical symptoms of disease with a low bacterial load.
In conclusion, our study is the first to our knowledge demonstrating that the optimal antibiotic dose of a fluoroquinolone was markedly different between two bacterial species that exhibit similar MIC and are closely genetically related. The appropriate and prudent use of antimicrobials is a growing challenge for both human and veterinary medicine, and as already suggested, the optimization of antibiotic dosing regimens should take all its place in this challenge. In this context, our present results could stimulate further research to investigate the potential implementation of bacterial species-specific antibiotic doses in a clinical setting.

Author Contributions
Conceived and designed the experiments: GL FEG PLT AAF ABM. Performed the experiments: GL FEG AAF. Analyzed the data: GL FEG PLT AAF ABM. Contributed reagents/materials/analysis tools: GL FEG PLT AAF ABM. Wrote the paper: GL FEG PLT AAF ABM.