The Dysbiosis Triggered by First-Line Tuberculosis Antibiotics Fails to Reduce Their Bioavailability

ABSTRACT Antituberculosis therapy (ATT) causes a rapid and distinct alteration in the composition of the intestinal microbiota that is long lasting in both mice and humans. This observation raised the question of whether such antibiotic-induced changes in the microbiome might affect the absorption or gut metabolism of the tuberculosis (TB) drugs themselves. To address this issue, we utilized a murine model of antibiotic-induced dysbiosis to assay the bioavailability of rifampicin, moxifloxacin, pyrazinamide, and isoniazid in mouse plasma over a period of 12 h following individual oral administration. We found that 4-week pretreatment with a regimen of isoniazid, rifampicin, and pyrazinamide (HRZ), a drug combination used clinically for ATT, failed to reduce the exposure of any of the four antibiotics assayed. Nevertheless, mice that received a pretreatment cocktail of the broad-spectrum antibiotics vancomycin, ampicllin, neomycin, and metronidazole (VANM), which are known to deplete the intestinal microbiota, displayed a significant decrease in the plasma concentration of rifampicin and moxifloxacin during the assay period, an observation that was validated in germfree animals. In contrast, no major effects were observed when similarly pretreated mice were exposed to pyrazinamide or isoniazid. Thus, the data from this animal model study indicate that the dysbiosis induced by HRZ does not reduce the bioavailability of the drugs themselves. Nevertheless, our observations suggest that more extreme alterations of the microbiota, such as those occurring in patients on broad-spectrum antibiotics, could directly or indirectly affect the exposure of important TB drugs and thereby potentially influencing treatment outcome.

IMPORTANCE Previous studies have shown that treatment of Mycobacterium tuberculosis infection with first-line antibiotics results in a long-lasting disruption of the host microbiota. Since the microbiome has been shown to influence the host availability of other drugs, we employed a mouse model to ask whether the dysbiosis resulting from either tuberculosis (TB) chemotherapy or a more aggressive course of broadspectrum antibiotics might influence the pharmacokinetics of the TB antibiotics themselves. While drug exposure was not reduced in animals previously described as exhibiting the dysbiosis triggered by conventional TB chemotherapy, we found that mice with other alterations in the microbiome, such as those triggered by more intensive antibiotic treatment, displayed decreased availability of rifampicin and moxifloxacin, which in turn could impact their efficacy. The above findings are relevant not only to TB but also to other bacterial infections treated with these two broader spectrum antibiotics. KEYWORDS  This is a work of the U.S. Government and is not subject to copyright protection in the United States. Foreign copyrights may apply.
T uberculosis (TB) continues to be a leading cause of death in many parts of the world (1). The current standard treatment for drug-sensitive clinical TB requires a 6month treatment regimen of daily oral administration of the antibiotics rifampicin, isoniazid, pyrazinamide, and ethambutol for 2 months and rifampicin and isoniazid for 4 months. While this antituberculosis therapy (ATT) is effective, the long duration of the treatment can lead to noncompliance promoting the emergence of drug resistance in mycobacteria (2). Furthermore, while many patients that complete the 6-month course of ATT are cured from active TB, certain patients require extended treatment or do not respond as well to treatment (3,4). Additionally, a proportion of those treated and cured can develop TB again either due to reinfection or reactivation of any dormant bacilli that persisted despite a full course of antibiotics (5). This variable outcome observed during and following ATT is not fully understood.
One of the factors that has emerged as a determinant of chemotherapeutic intervention efficacy is the intestinal microbiota (6). Studies in the last decade have demonstrated the complex and bidirectional influence and interaction between the gut microbiome and several nonantibiotic drugs, such as metformin, digoxin, levodopa, serotonin uptake inhibitors, proton pump inhibitors, and immune checkpoint blockades (7)(8)(9)(10)(11)(12). While in certain cases drug treatment can affect the composition of the gut flora, the microbes contained within may in turn enzymatically affect the drugs themselves, influencing their bioactivity, availability, and toxicity positively or negatively (12).
With regard to the microbiota in the setting of TB and ATT, we and others have shown that ATT results in a profound alteration of the intestinal microbiota in both patients and animal models and that this dysbiosis persists long after the cessation of treatment (13)(14)(15)(16). Using our mouse model, we observed that these changes are firmly established as early as 4 weeks after the initiation of antibiotic treatment (13). The latter observations prompted the question of whether such pronounced effects on the microbiota might influence the bioavailability or exposure of the antibiotics themselves. In previous studies using murine and rhesus macaque models, we and others established that Mycobacterium tuberculosis infection by itself does not result in a dramatic alteration of the microbiota and that the changes occurring are subtle in comparison to the effects of ATT (17)(18)(19). Thus, in addressing the possible influence of the microbiota on drug dynamics, we utilized a murine model of 4-week antibiotic treatment in the absence of M. tuberculosis infection itself. To this end, one group of mice was treated orally with a combination of front-line TB drugs, namely, isoniazid, rifampicin, and pyrazinamide (HRZ), for 26 days. A second animal group received a cocktail of the antibiotics vancomycin, ampicillin, neomycin. and metronidazole (VANM) for the same duration. We have previously characterized and reported the highly reproducible compositional changes induced by these two antibiotic regimens in the murine intestinal microbiota (13). VANM treatment is routinely used in microbiome-based studies to deplete the gut microbiota, and this group was included to examine the overall importance of the microbiota in drug exposure in an antibiotic administration model (20). We also included a third group consisting of germfree (GF) animals to directly investigate the importance of the microbiota in drug exposure and confirm the results of the VANM group (Fig. 1A).
Two days following the cessation of antibiotic pretreatment, each group of mice received an oral dose of rifampicin. The animals were then serially bled over a 12-h time course and the antibiotic concentrations were measured in plasma by high-pressure liquid chromatography coupled to tandem mass spectrometry (HPLC/MS-MS). We observed significantly higher levels of rifampicin in the plasma of HRZ-treated versus control (Ctrl) mice not previously treated with antibiotics, at the early time points of 1 and 2 h following rifampicin administration. However, rifampicin was cleared faster from the circulation in HRZ-treated animals as we detected significantly lower levels in these mice than those of controls at 12 h (Fig. 1B). As a result, no significant difference in drug exposure was observed between these two groups in a 12-h period, as measured by the area under the concentration-time curve (Fig. 1C). Interestingly, animals that received the VANM (instead of HRZ) cocktail displayed a lower plasma concentration of rifampicin at several time points, and the overall antibiotic exposure was significantly lower than that of untreated controls or HRZ-treated animals ( Fig. 1B and C). Furthermore, germfree animals displayed a phenotype similar to that of VANM-treated animals, albeit with higher plasma levels of rifampicin in the first hour after administration ( Fig. 1B and C). Together, these data suggest that the microbiota and prior antibiotic treatment history can affect the pharmacokinetics and bioavailability of rifampicin in mice.
In order to investigate the importance of the microbiota in the optimal bioavailability of other antibiotics, we performed a similar pharmacokinetic (PK) study with the broad-spectrum and second-line TB antibiotic moxifloxacin and two front-line mycobacterium-specific TB drugs, namely, pyrazinamide, and isoniazid. Moxifloxacin displayed a PK profile like that of rifampicin, that is, in comparison to untreated animals, moxifloxacin exposure in a 12-h period was not different in HRZ-treated animals, while it was significantly lower in VANM-treated animals ( Fig. 2A and B). Interestingly, in the case of pyrazinamide and isoniazid, VANM treatment significantly lowered pyrazinamide exposure only compared with HRZ-treated animals, and while isoniazid exposure trended lower in the VANM group in the same comparison, this difference was not significant ( Fig. 2C and D; see Fig. S1A and B in the supplemental material).
Taken together, these findings indicate that the microbiota can directly or indirectly affect the pharmacokinetics of some but not all oral TB antibiotics. In particular, our data reveal that prior frontline ATT with HRZ fails to reduce the bioavailability of The plasma concentration of rifampicin for the four experimental groups is plotted over time. Statistical significance between two groups for each time point was calculated using the Mann-Whitney U test, and two groups that are significantly different are represented using a different symbol and color as denoted in the legend. (one symbol, P , 0.05; two symbols, P , 0.01; three symbols, P , 0.001; four symbols, P , 0.0001) (C) Area under the concentration-time (AUC) curve from 0 to 12 h was calculated for each group. One-way analysis of variance (ANOVA) was used to calculate statistical significance (**, P , 0.01; ***, P , 0.001; ****, P , 0.0001). n = 5 and data are pooled from two independent experiments. rifampicin and moxifloxacin, suggesting that HRZ-induced dysbiosis is unlikely to be an important determinant of the interindividual variable responses against M. tuberculosis encountered during treatment with this regimen. In contrast, we found that the total or near-complete elimination of the microbiome occurring in germfree or VANMtreated mice significantly lowers exposure of these two TB antibiotics. While not explored in the current study, possible mechanisms for this positive function of the microbiota on rifampicin-moxifloxacin absorption include altered first-pass metabolism in the liver, effects on passive transport through the intestinal wall, or alterations in the catalytic reduction of the antibiotic. Our unexpected finding that HRZ treatment can result in transiently higher rifampicin exposure may relate to the complex induction and inhibition patterns of transporters and drug-metabolizing enzymes by rifampicin, which in mice have yet to be fully elucidated (21). Interestingly, isoniazid is a prodrug that requires catalase from M. tuberculosis for activation, and the antibacterial drug Prontosil requires bacterial azoreductases for its antibacterial activity (22,23). Similar bacterial influences on rifampicin or moxifloxacin function have not been described but could potentially explain the effects of the microbiota on other antibiotics used in TB treatment.
Our observations are the first to document the association of an intact microbiome with the optimal PK of TB antibiotics but require further investigation to define the mechanisms involved. In particular, microbiota restoration and colonization studies are needed to confirm the direct role of the microbiome (as opposed to possible physiologic changes triggered by prior antibiotic administration) in mediating the effects observed. If appropriately followed up, our findings could be clinically relevant for TB patients treated previously with broad-spectrum antibiotics or with other prolonged antibiotic regimens severely affecting the gut flora. Indeed, recent data advocating the use of moxifloxacin in a 4-month TB treatment regimen make these observations particularly relevant as one would predict a more severe disruption of the microbiome with that combination regimen given the broad-spectrum efficacy of moxifloxacin (24). Thus, the possibility remains that other current or future TB antibiotic regimens distinct from HRZ may also alter the PK of the drugs employed and that such effects could impact their efficacy against M. tuberculosis infection. Finally, our observations on the effects of prior broad-spectrum antibiotic administration on rifampicin or moxifloxacin availability may be of relevance in the treatment of other bacterial diseases with these drugs. Methods. One group of specific-pathogen-free C57BL/6 mice was gavaged orally 5 days a week for 26 days with 200 mL of a combination of isoniazid (25 mg/kg of body weight), rifampicin (10 mg/kg), and pyrazinamide (150 mg/kg). Another group of mice received an oral antibiotic cocktail of vancomycin (10 mg/kg), ampicillin (20 mg/kg), neomycin (20 mg/kg), and metronidazole (20 mg/kg) for the same period (all from Sigma-Aldrich). The treatment was stopped after 26 days. Two days after cessation of antibiotic administration, mice were fasted overnight in preparation for the PK studies, except in the case of the isoniazid PK study where the mice were not fasted. These antibiotic-pretreated mice as well as groups of untreated and C57BL/6 mice reared under germfree conditions were then given a test dose of rifampicin (10 mg/kg), moxifloxacin (100 mg/kg), pyrazinamide (150 mg/kg), or isoniazid (25 mg/kg) administered by oral gavage and were serially bled at 30 min, 1 h, 2 h, 6 h, and 12 h postgavage, except in the case of the isoniazid study. Food was provided after the 2-h blood collection time point. GF mice were similarly fasted overnight and maintained in an isolator until oral administration of an antibiotic for PK measurement wherein they were moved to a biosafety cabinet for the duration of the experiment. Each animal group consisted of at least 5 female mice aged 6 to 8 weeks. All experimental procedures were in compliance with protocols approved by the NIAID/NIH Animal Care and Use Committee. Plasma samples collected were stored in 280°C and shipped on dry ice for analysis using HPLC/MS-MS as described previously (25).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. FIG S1, PDF file, 0.4 MB.

ACKNOWLEDGMENTS
This work was supported by the Intramural Research Program of the NIAID and NIH grant number S10-OD023524 awarded to V.D.
Germfree mice were provided by the NIAID gnotobiotic animal facility with the generous assistance of Nicolas Bouladoux. We also thank the NIAID Building 50 Animal Facility Staff for performing daily oral gavage of the antibiotics employed in this study.
We have declared that no conflict of interest exists.