ABSTRACT
Antibiotic susceptibility testing (AST) in Neisseria meningitidis is an important part of the management of invasive meningococcal disease. It defines MICs of antibiotics that are used in treatment and/or prophylaxis and that mainly belong to the beta-lactams. The interpretation of the AST results requires breakpoints to classify the isolates into susceptible, intermediate, or resistant. The resistance to penicillin G is defined by a MIC of >0.25 mg/liter, and that of amoxicillin is defined by a MIC of >1 mg/liter. We provide data that may support revision of resistance breakpoints for beta-lactams in meningococci. We used experimental intraperitoneal infection in 8-week-old transgenic female mice expressing human transferrin and human factor H. Dynamic bioluminescence imaging was performed to follow the infection by bioluminescent meningococcus strains with different MICs. Three hours later, infected mice were treated intramuscularly using several doses of amoxicillin or penicillin G. Signal decreased during infection with a meningococcus strain showing a penicillin G MIC of 0.064 mg/liter at all doses. Signals decreased for the strain with a penicillin G MIC of 0.5 mg/liter only after treatment with the highest doses, corresponding to 250,000 units/kg of penicillin G or 200 mg/kg of amoxicillin, although this decrease was at a lower rate than that of the strain with a MIC of 0.064 mg/liter. The decrease in bioluminescent signals was associated with a decrease in the levels of the proinflammatory cytokine interleukin-6 (IL-6). Our data suggest that a high dose of amoxicillin or penicillin G can reduce growth during infection by isolates showing penicillin G MICs of >0.25 mg/liter and ≤1 mg/liter.
INTRODUCTION
Neisseria meningitidis is a Gram-negative bacterium frequently encountered in the human nasopharynx, but it is also the causative agent of invasive meningococcal disease (IMD) that provokes mainly septicemia and meningitis. Neisseria meningitidis remains susceptible to beta-lactams, the antibiotics of choice in the treatment of IMD (1). Resistance to beta-lactams in meningococci is extremely rare, but reduced susceptibility to penicillin G and to amoxicillin (intermediate isolates; Peni) has been described previously. However, neither resistance nor reduced susceptibility to third-generation cephalosporins has been detected so far (2). The proportions of Peni isolates differ worldwide and are increasing in several countries and can reach >30% of total meningococcal isolates (3–7).
We have previously shown a direct correlation between the polymorphism of the penA gene encoding the penicillin binding protein 2 (PBP2) and the Peni phenotype. This phenotype seems to result from the reduced affinity of penicillin G and amoxicillin to PBP2 as well as to modification of the peptidoglycan structure in Peni isolates with increased pentapeptide-containing muropeptides (8). Horizontal interspecies DNA exchanges in the genus Neisseria are suggested to drive the polymorphism of penA (7). Antibiotic susceptibility testing (AST) is mandatory for beta-lactam antibiotics and requires reliable breakpoints to inform decision making in patient treatment.
In order to consistently define breakpoints, penA genes from a large collection of isolates were sequenced, and this allowed the linking of wild-type alleles of penA to a low MIC for penicillin G (<0.125 mg/liter) (7). This defined the epidemiological susceptibility cutoff values for the MIC of penicillin G to be lower than 0.125 mg/liter and that of amoxicillin to be <0.250 mg/liter(7). It divided the meningococcal population into one part containing isolates harboring wild-type alleles of penA and another part comprising isolates showing highly diverse penA alleles and MICs of ≥0.125 mg/liter and 0.250 mg/liter for penicillin G and amoxicillin, respectively (7). The MIC value of <0.125 mg/liter was preferred to define the susceptibility to penicillin G as it fulfills the important rule of not splitting wild-type MIC distributions (9) as isolates with wild-type penA genes showed a MIC of 0.094 mg/liter (7). These values are consistent with those used by the European Committee for Antimicrobial Susceptibility Testing (EUCAST [http://www.eucast.org]) and the Clinical and Laboratory Standards Institute (CLSI) (10).
Intermediate isolates are expected to be treatable by beta-lactams; i.e., bacteria growth is reduced, and/or bacteria are cleared from biological fluids. However, the higher limit of Peni isolates is still to be determined. EUCAST and CLSI indicate that isolates with penicillin G and amoxicillin MICs of >0.250 mg/liter and >1 mg/liter, respectively, are resistant (Penr) to these beta-lactams (i.e., nontreatable/treatment failure). However, isolates with a penicillin G MIC of >0.250 mg/liter harbor penA alleles similarly modified as those of Peni isolates (7). The definition of resistance breakpoints is mainly driven by pharmacokinetic (PK) and pharmacodynamic (PD) indices that reflect the concentration and effect, respectively, of the antibiotic. However, experimental data are needed to correlate breakpoints with treatment. The use of animal models may help test whether these breakpoints correspond to resistance and treatment failure.
MATERIALS AND METHODS
Ethics statement.This study was carried out in strict accordance with the European Union Directive 2010/63/EU (and its revision, 86/609/EEC) on the protection of animals used for scientific purposes. Our laboratory has administrative authorization for animal experimentation (permit number 75-1554), and the protocol was approved by the Institut Pasteur Review Board that is part of the Regional Committee of Ethics of Animal Experiments of Paris region (permit number 99-174).
Meningococcal isolates: phenotypic and genotypic characterization.Two clinical isolates of N. meningitidis were used (LNP24198 and LNP27704). Both isolates were of serogroup C and belonged to the clonal complex ST-11 (cc11). They harbored, respectively, the penA alleles penA3 and penA9 corresponding to wild-type and modified alleles, respectively. The MICs of penicillin G were determined as previously recommended using Etest with Mueller-Hinton agar supplemented with sheep blood (11) and were 0.064 mg/liter and 0.5 mg/liter, respectively. MICs of amoxicillin were 0.125 and 1.5 mg/liter, respectively. Bioluminescent variants of both isolates were constructed by transformation with the recombinant plasmid pDG34, which carries the bioluminescent luxCDABE operon under the control of the porB promoter (12), and named LNP24198lux and LNP27704lux. Both strains were checked for their MICs of penicillin G and amoxicillin, and their penA alleles were verified by sequencing, showing that they were identical to those of the parent isolates. Both strains grew similarly on meningococcal growth medium.
Mouse infection and dynamic live-imaging studies.To study N. meningitidis infection, we took advantage of the availability of an animal model, the transgenic mouse expressing the human transferrin, since an iron source is required for meningococcal growth (13). We have recently developed another animal model, a transgenic mouse expressing the human factor H (fH) that allowed binding of this negative regulator of the complement pathway to the bacterial surface and hence allowed meningococci to escape complement-mediated lysis (14). The two types of mice were crossed to generate transgenic mice expressing both human transferrin and human fH that we used in infection experiments using bioluminescent meningococcal strains with different penicillin G and amoxicillin MICs. Mice were bred in-house and were kept in a biosafety containment facility in filter-topped cages with sterile litter, water, and food, according to institutional guidelines.
Mice were infected by the intraperitoneal route (i.p.) with a standardized inoculate of 5 × 106 bioluminescent CFU per mouse in 0.5 ml of bacterial suspension. At the time point of 3 h, the mice were divided into three groups that were treated by either penicillin G or amoxicillin only once by intramuscular injection in the interior face of the left thigh. The following increasing unique doses (per mouse) were used: penicillin G at 60,000 units/kg, 120,000 units/kg, or 250,000 units/kg, corresponding to 37 mg/kg, 75 mg/kg, and 150 mg/ml, or amoxicillin at 50 mg/kg, 100 mg/kg, and 200 mg/kg. The highest doses of both antibiotics corresponded to a daily dose used in treatment of IMD in humans. A group of two mice was injected only with saline as a control. Interleukin-6 (IL-6) levels were also assayed in blood at the endpoint (8 h), as previously described (15). As similar results were obtained from both experiments with penicillin G and amoxicillin, IL-6 was tested in mice treated with amoxicillin.
Bacterial infection images were acquired after 0.5 h, 3 h, 6 h, and 8 h of infection using an IVIS 100 system (Xenogen Corp., Alameda, CA) as previously described (13). Analysis and acquisition were performed using Living Image, version 4.3.1, software (Xenogen Corp.). Data were analyzed by linear regression using GraphPad InStat, version 3.06 (GraphPad Software, San Diego, CA, USA).
RESULTS
Impact of MIC on amoxicillin treatment during meningococcal infection in mice.In order to test the evolution of the experimental infection in vivo, transgenic mice were infected i.p. with the LNP24198lux (penicillin MIC, 0.064 mg/liter; amoxicillin MIC, 0.125 mg/liter) or LNP27704lux (penicillin MIC, 0.5 mg/liter; amoxicillin MIC, 1.5 mg/liter) strain. After 3 h of infection, infected mice were divided into groups that were treated by one of the three antibiotic doses (amoxicillin or penicillin G). A group of mice was left untreated as a control group.
After bacterial intraperitoneal challenge, dynamic bioluminescence imaging showed that 30 min after the bacterial suspension injection, bacteria were present mainly in the peritoneal cavity in mice infected with either strain LNP24198lux or strain LNP27704lux. Signal increased in all mice after 3 h of infection. For mice infected with strain LNP24198lux, 6 h after infection (3 h after treatment with amoxicillin), the signal decreased in mice treated with a 200-mg/kg dose, and bacteria were cleared after 8 h of infection (5 h after treatment). For the two other doses, the signal decreased only after 8 h of infection (5 h after treatment). In the untreated mice, signal continued to increase at all time points (Fig. 1A). For mice infected with strain LNP27704lux (penicillin MIC, 0.5 mg/liter; amoxicillin MIC, 1.5 mg/liter), the bioluminescent signal decreased only in mice treated with the highest dose (200 mg/kg) while signal continued to increase in mice treated with 50 mg/kg and 100 mg/kg and did not differ from the signal observed in untreated mice (Fig. 1B).
Transgenic mice were infected i.p. with a standardized inoculum of 5 × 106 bioluminescent CFU per mouse in 0.5 ml of bacterial suspension. Quantification was performed after 30 min, 3 h, 6 h, and 8 h of infection by defining regions of interest (the whole mouse). After 3 h of infection, mice were treated with the unique indicated doses of amoxicillin or penicillin G by intramuscular injection in the interior face of the left thigh. The untreated mice received an injection with the same volume of physiological serum. Bioluminescent images are shown at left, and linear regression analyses of the evolution of bioluminescent signals after treatment are shown at right. Linear regression data are shown as solid lines with colors corresponding to the indicated tested condition, with dashed color-matched lines corresponding to the 95% confidence band. (A) Amoxicillin treatment data for the strain LNP24198lux (MICs of 0.064 mg/liter and 0.125 mg/liter of penicillin G and amoxicillin, respectively). (B and C) Amoxicillin treatment and penicillin G treatment data, respectively, for strain LNP27704lux (MICs of 0.5 mg/liter and 1.5 mg/liter of penicillin G and amoxicillin, respectively). The doses of amoxicillin (A and B) and those of penicillin G (C) are shown above the images. Radiance is measured as photons per second per square centimeter per steradian.
Linear regression analysis confirmed the significant decrease of bacterial viability for the strain LNP24198lux in amoxicillin-treated mice in a dose-dependent manner. Negative slopes were also significantly different from the slope for untreated mice (P < 0.01 and r2 of 0.38 and 0.85, respectively, for 100-mg/kg and 200-mg/kg doses) (Fig. 1A and B). However, for the dose of 50 mg/kg, the slope was not significantly different from zero (P = 0.77 and r2 = 0.007), suggesting a bacteriostatic effect (Fig. 1A and B).
For strain LNP27704lux, only the 200-mg/kg dose resulted in a negative slope that differed significantly from that observed with untreated mice and from the slopes for mice treated with the other doses of amoxicillin (P < 0.001 and r2 = 0.59). For the mice infected with strain LNP27704lux and treated with the two other doses (50 mg/kg and 100 mg/kg), the slopes were positive and similar to those of untreated mice, suggesting bacterial growth (Fig. 1A and B). For the 200-mg/kg-treated mice, the r2 value for infection with strain LNP27704lux was lower than that with strain LNP24198lux (0.59 versus 0.85, respectively), suggesting slower clearance of the former strain.
We also performed similar experiments using penicillin G with doses of 60,000 units/kg, 125,000 units/kg, and 250,000 units/kg. Mice were infected i.p. with strain LNP27704 (penicillin MIC, 0.5 mg/liter; amoxicillin MIC, 1.5 mg/liter), and after 3 h of infection, three groups of infected mice were treated by one of the three penicillin G doses. A group of mice was left untreated as a control.
Bioluminescent signals increased and spread after 3 h of infection. After treatment, only the group of mice treated with the highest dose (250,000 units/kg) showed a decrease in bioluminescent signal indicating loss of bacterial growth. The signal continued to increase in mice that were infected with LNP27704lux and treated with either 60,000 units/kg or 125,000 units/kg of penicillin G (Fig. 1C).
Linear regression analysis confirmed the impact of the highest dose in mice infected with strain LNP27704lux. The highest dose resulted in a negative slope that differed significantly from that obtained with the other doses and from that of untreated mice (P < 0.001), with an r2 value of 0.12.
For the mice infected with strain LNP24198lux (MIC of 0.064 mg/liter) and treated with penicillin G, the data were similar to those obtained with amoxicillin treatment (data not shown).
All these data taken together suggest that the highest dose, similar to the daily dose used in humans, leads to a decrease in bacterial counts in mice even if the strain is considered Penr.
Decreased inflammatory response in mice treated with an optimal amoxicillin dose.We next assayed the levels of IL-6, a proinflammatory cytokine, at 8 h postinfection. The data are shown in Table 1 and clearly suggest a dose-dependent decrease of IL-6 levels in amoxicillin-treated mice that were infected with strain LNP24198lux (penicillin MIC, 0.064 mg/liter; amoxicillin MIC, 0.125 mg/liter). These levels were not detectable in mice treated with 200 mg/kg of amoxicillin. On the other hand, mice infected with strain LNP27704lux (penicillin MIC, 0.5 mg/liter; amoxicillin MIC, 1.5 mg/liter) and treated with 50 mg/kg and 100 mg/kg of amoxicillin showed similar levels of IL-6 as those obtained in untreated mice (Table 1). IL-6 was undetectable in mice treated with the highest dose (200 mg/kg of amoxicillin).
Il-6 levels in the mice shown in Fig. 1 after 8 h of infection
DISCUSSION
We have previously emphasized the advantages of in vivo bioluminescent imaging for real-time monitoring of meningococcal infections and their treatment (13, 16). This technology is sensitive, rapid, and noninvasive. It limits animal-to-animal variation and reduces the number of animals used. In the present study, we adapted this system to evaluate the correlation of beta-lactam breakpoints during experimental meningococcal infection in transgenic mice. N. meningitidis is intrinsically susceptible to many antibiotic classes and MIC50 and MIC90 values (antibiotic concentrations, respectively, inhibiting 50% and 90% of a sample of strains) are very low (17). Meningococcal resistance to beta-lactams remains rare, and meningococcal strains remained susceptible to penicillin G, with a penicillin G MIC50 of 0.06 mg/liter. Few previous reports describe rare beta-lactamase-producing strains with MICs that may reach or exceed 256 μg/ml (18, 19). These beta-lactamases are plasmid-borne TEM-1-type enzymes, similar to those described in Neisseria gonorrhoeae, that inactivate penicillin G and amoxicillin but do not hydrolyze third-generation cephalosporins (18). However, isolates with reduced susceptibility to penicillin G and amoxicillin are quite frequent worldwide (7). The MICs of amoxicillin and penicillin G for these intermediate isolates remain low, and defining the upper limit of the critical values remains problematic.
Although treatment failures have been described for strains with the highest MICs (20), the severe infections caused by these strains generally resolve favorably using high doses of penicillin G or amoxicillin, which allow bactericidal concentrations to be reached in cerebrospinal fluid (CSF). The effect of beta-lactams is dependent on the time (T) the antibiotic concentration exceeds its MIC for the microorganism (T > MIC) (9). Our data with the highest dose of antibiotics are in agreement with this consideration. This dose corresponds to the recommended total daily dose for the treatment of IMD (21). Our data suggest that the strain with a MIC of 0.5 mg/liter may not be classified as resistant to penicillin G as the infection was treatable. However, lower doses of penicillin G and amoxicillin failed to control bacterial growth and might lead to treatment failure. As the threshold of 1 mg/liter corresponds to the effective therapeutic concentration in the CSF obtained during treatment with penicillin G (22), we suggest that the upper breakpoint for penicillin G be 1 mg/liter, and thus intermediate isolates may be defined as those with MICs of ≥0.125 mg/liter and ≤1 mg/liter. This range also contains the isolates with altered penA alleles that all shared the same altered residues (7). Isolates with MICs of >0.25 mg/ml and ≤1 mg/ml may not be considered Penr isolates. Indeed, the drug effect, in vivo, is subject to the bacterial growth rate but is also dependent on host defense mechanisms. The latter aspects are not directly considered by the conventional PK/PD model but can be addressed using relevant animal models, as described in this work.
IL-6 is considered to be an important mediator of acute inflammatory responses to bacterial infection (23). After 8 h of infection, the levels of IL-6 increased in mice infected by both isolates and were higher in mice infected by the susceptible strain (LNP24198lux) than in mice infected by the resistant isolate (LNP27704lux). This is in agreement with our previous study that meningococcal isolates with a modified PBP2 showed significantly lower induction of the inflammatory response (15). The mice treated with the higher dose did not show any detectable IL-6. The trends of IL-6 levels in amoxicillin-treated groups followed the microbiological results. Our data are consistent with previous reports suggesting that IL-6 is a possible indicator of bacterial killing (24).
Finally, our data suggest that if amoxicillin and penicillin G are to be used in treatment of IMD prior to determination of the MIC, the first doses should be 200 mg/kg and 250,000 units/kg, respectively.
Our findings clearly highlight a powerful approach in defining breakpoints through the analysis the polymorphism of the gene encoding the targets of the antibiotics to characterize the alterations of the targets and their correlation with the MIC. The use of animal models in association with PK/PD analysis can then allow determination of the breakpoints. Finally, the increase of the MIC in meningococci may also correspond to other undefined mechanisms that require additional studies. However, our proposal for a breakpoint for resistance to penicillin G higher than 1 mg/liter fits with other breakpoints for pathogens showing meningeal tropism (25).
ACKNOWLEDGMENTS
The work was supported by the Institut Pasteur.
This work used the dynamic imaging facilities of the Imagopole at the Institut Pasteur, Paris, France.
FOOTNOTES
- Received 16 February 2016.
- Returned for modification 20 March 2016.
- Accepted 14 April 2016.
- Accepted manuscript posted online 18 April 2016.
- Copyright © 2016, American Society for Microbiology. All Rights Reserved.
