Toward a Bactericidal

Oral Drug Combination for the Treatment of Mycobacterium abscessus Lung Disease

Authors: Jickky Palmae Sarathy, Min Xie, Chui Fann Wong, Dereje A. Negatu, Suyapa Rodriguez, Matthew D. Zimmerman, Diana C. Jimenez, Ilham M. Alshiraihi, Mercedes Gonzalez-Juarrero, Véronique Dartois, Thomas Dick

Abstract

Treatment of Mycobacterium abscessus lung disease relies on underperforming drug combinations and includes parenteral, poorly tolerated, and bacteriostatic antibiotics. We posit that safe, oral, and bactericidal regimens are needed to improve cure rates and shorten treatment. Here, we combined oral representatives of three well-tolerated bactericidal drug classes, the β-lactam tebipenem (together with the β-lactamase inhibitor avibactam), the fluoroquinolone moxifloxacin, and the rifamycin rifabutin, and profiled the combination in vitro and in vivo. The combination potentiated bactericidal activity of its components against replicating M. abscessus and retained bactericidal activity against the nonreplicating, drug-tolerant form of the bacterium residing in surrogate caseum. When combined, the drugs retained the ability to induce lethal secondary effects associated with the β-lactam and fluoroquinolone, including cell wall and DNA damage, increased metabolism, and generation of reactive oxygen species. Thus, the triple-drug combination appears to exert two lethal punches while suppressing bacterial reprogramming to counter the drug-induced stresses, providing a plausible rationale for the enhanced kill effect. Addition of a bacteriostatic agent resulted in drug-specific patterns of interactions with regards to bactericidal activity reflected by the lethal secondary effects. The triple-drug combination also exerted a pronounced postantibiotic effect and reduced emergence of spontaneous resistant mutants. Collectively, this work provides a combination prototype for optimization and a profiling workflow that may be useful for the development of sterilizing regimens.

Introduction

Mycobacterium abscessus is a fast-growing environmental saprophyte that can cause pulmonary disease similar to its slow growing relative Mycobacterium tuberculosis. In contrast to the obligate human pathogen, M. abscessus is an opportunistic pathogen that typically affects patients with structural lung damage and/or immunodeficiencies.^1^ Also, in contrast to tuberculosis (TB), for which curative regimens are available, no reliable curative treatment is available for M. abscessus disease.^2^ Optimal therapeutic strategies, including drug selection and treatment duration, are unknown. There is no FDA-approved antibiotic for the treatment of M. abscessus lung infections. Guidelines suggest a biphasic approach with each phase including at least three and two drugs, respectively.^3^ An initial intensive phase of 3 to 12 weeks includes parenteral agents and is followed by a continuation phase that comprises oral agents, usually in combination with inhaled amikacin. Treatment can take six months to a year or longer, followed by 12 months of treatment after culture conversion is achieved to prevent relapse. Injectable agents include amikacin, a β-lactam (imipenem or cefoxitin), and tigecycline. Oral agents include a macrolide (clarithromycin (CLR) or azithromycin) as cornerstone, clofazimine, linezolid, and moxifloxacin (MXF). More recently, omadacycline, tedizolid, bedaquiline (BDQ), and rifabutin (RFB) have been considered as oral add-on options.^4^ Despite prolonged multiphase and multidrug treatment, cure rates are unacceptably low (∼35% for disease caused by strains displaying inducible macrolide resistance).^5^ In addition, current regimens are associated with serious adverse events. Thus, there is an urgent medical need for all-oral, safer, and, importantly, more efficacious treatment regimens. Despite major biological uncertainties around desirable properties of drugs or drug classes to combine,^6^ a large-scale clinical study recently showed that antibiotic lethality dictates mycobacterial infection outcomes.^7^

Among the factors driving the long and ineffective treatment is the lack of a competent immune response in a large fraction of patients, associated with limited access of the immune system to the pathogen occupying anatomical niches (e.g., necrotic lung lesions or mucus in airways of cystic fibrosis patients), drug-induced immunosuppression (e.g., in organ transplantation or cancer chemotherapy), or aging. This is different from other bacterial infections where antibiotics work in tandem with the immune system to eradicate the pathogen, and bacteriostatic drugs can be relied on despite only suppressing proliferation of the microbe. Therefore, designing drug regimens with strong sterilizing activity appears as a sensible strategy to cure M. abscessus pulmonary disease, consistent with recent clinical data.^7^

We posit that sterilizing activity against M. abscessus can be achieved by combining drugs that exert bactericidal activity against both replicating and nonreplicating bacilli. This notion is supported by the experience gained in TB, a disease which displays pathophysiology similarities with M. abscessus lung ^8^ treatment-shortening TB regimens include drugs that kill both forms of the tubercle bacillus. Given that nonreplicating mycobacteria, which typically reside in the necrotic center (caseum) of lung granulomas, exhibit extreme tolerance to various bactericidal antibiotics,^9,10^ the ability to kill this physiological form appears to be paramount to the sterilizing activity of antimycobacterial drug regimens.

Recent work has identified members of the β-lactam, fluoroquinolone, and rifamycin classes that exert bactericidal activity against nonreplicating M. abscessus in caseum, in addition to being bactericidal against replicating bacteria, which contrasts with the weak bactericidal activities exerted by most standard-of-care anti-M. abscessus drugs, including CLR.^10^ Hence, we hypothesized that combining a β-lactam, fluoroquinolone, and rifamycin may enhance the bactericidal activities of the component drugs and thus generate attractive efficacy. The selection of these three drug classes for combination studies against M. abscessus is supported by clinical experience gained from TB. Fluoroquinolones and rifamycins play key roles in treatment shortening of TB.^11^ The cell-wall synthesis inhibitor isoniazid is a rapid bactericidal first-line drug in TB treatment but is not active against M. abscessus and therefore replaced with a β-lactam, also bactericidal and targeting cell-wall biosynthesis.

Studies of the kill mechanisms of bactericidal antibiotics have revealed that their primary mechanism of action, specifically the inhibition of their molecular target, triggers intrabacterial secondary events which are critical or at least contribute to the lethal effect of the drugs.^12^ Hence, bacterial cell death is caused by a combination of primary target corruption and secondary events. The intrabacterial events triggered by bactericidal antibiotics have been extensively studied in Gram-negative and -positive bacteria, and to some extent in M. tuberculosis, but less in M. abscessus. β-Lactams covalently modify penicillin-binding proteins, eliminating their transpeptidase and peptidoglycan cross-linking activity, while peptidoglycan hydrolase activities continue, making them suicide inhibitors. The ensuing imbalance between the synthesis and degradation of the peptidoglycan layer of the cell wall results in cell lysis. Fluoroquinolones covalently trap DNA within the gyrase complex, preventing the enzyme from resealing double-stranded DNA breaks, which can result in DNA fragmentation. Cellular repair responses to cell wall and DNA damage caused by β-lactams and fluoroquinolones, respectively, are energy costly. Hence, metabolic hyperdrive is triggered, causing increased generation of reactive oxygen species (ROS) and contributing to bacterial cell death.^13^ Rifamycins inhibit the RNA polymerase, suppressing transcription. We speculated that combining a β-lactam and a fluoroquinolone with a rifamycin may deliver a one-two punch to the structural integrity of the cell (i.e., disrupting cell wall and chromosome integrity) and induce lethal downstream metabolic changes (metabolic hyperdrive and ROS generation), while interfering with bacterial transcriptional rescue responses.^14^

In this work, we tested these concepts and hypotheses and evaluated the potential of the three-drug backbone for the development of an all-oral, safe, and efficacious drug regimen for M. abscessus pulmonary disease. Tebipenem (TBP), when used in combination with the β-lactamase inhibitor avibactam (TBP-A), was selected as the representative β-lactam as it was recently shown to be efficacious in a mouse model of M. abscessus lung infection and is more potent than the standard of care β-lactams imipenem and cefoxitin in vitro.^15,16^ Oral prodrugs of TBP and avibactam are in clinical development for other disease indications (ClinicalTrials.gov identifier NCT03788967 and NCT03931876, respectively), indicating their potential as repurposed oral drugs for use against M. abscessus. MXF and RFB were selected as representatives of the fluoroquinolone and rifamycin drug class, respectively, as they are currently considered as oral add-on options for the treatment of patients.^4^ In addition, these three drug classes do not suffer from significantly reduced potency against nonreplicating bacilli that reside in a caseum-like environment, unlike most other antibiotic classes.^10^

To evaluate the sterilizing potential of the TBP-A

• MXF + RFB oral backbone, their effect on growth inhibition, and bactericidal activity against replicating and nonreplicating bacteria, was determined. To gain mechanistic insights into their lethal secondary effects, alone and in combination, cell wall and DNA damage, metabolic activity, and ROS generation were quantified, including the impact of adding bacteriostatic antibiotics to the combination on these readouts. The postantibiotic effect (PAE) of the combination and the combination’s ability to reduce the emergence of resistance was determined. Collectively, the results feature TBP-A + MXF + RFB as a lethal combination prototype, forming the rational basis for the pharmacological optimization of its components against M. abscessus. The concepts and assay cascade, models, and readouts provide a framework for the preclinical evaluation, prioritization, and optimization of regimens for the treatment of M. abscessus lung disease.

Results

Dual- and Triple-Drug Combinations of TBP-A, MXF, and RFB Are

Largely Additive with Regards to Growth Inhibition

The aim of this study was to identify bactericidal drugs that, when combined, potentiate or at least do not antagonize their bactericidal activity. To ensure that the study drugs do not antagonize each other’s growth inhibitory activity, dual- and triple-drug combinations of TBP, MXF, and RFB were assessed via checkerboard analyses against M. abscessus ATCC 19977. TBP requires the β-lactamase inhibitor avibactam to exert potent activity against M. abscessus (Table S1).^16^ Thus, in this and all subsequent experiments involving TBP, avibactam was added at a fixed concentration of 4 μg/mL, abbreviated as TBP-A. Minimum inhibitory concentration (MIC; defined as the lowest concentration that inhibits 90% of bacterial growth compared to drug-free culture) was used to assess growth inhibition. The fractional inhibitory concentration index (FICI) was calculated to characterize the interaction between the drugs in different combinations. All drug pairs displayed additivity and the TBP-A + MXF + RFB combination was weakly synergistic (Table 1 and Data set S1). Importantly, their combinations did not reveal any antagonism.

Bactericidal Activities of TBP-A, MXF, and RFB Are Strongly

Potentiated When Combined

To determine the effect of combining TBP-A, MXF, and RFB on bactericidal activity, dual- and triple-drug combinations were assessed in time–concentration kill kinetic experiments against M. abscessus ATCC 19977. Drug treatment was carried out in liquid cultures at multiples of the respective MICs (TBP-A: 5 μM, MXF: 3 μM, RFB: 2.5 μM; Table S1), and bactericidal activity was quantified by CFU enumeration on a solid medium. As expected, the single drugs showed appreciable bactericidal activity, even at their MICs (∼0.5 log10 CFU/mL reduction), consistent with previous reports.^10,16,17^ At 2× MIC, TBP-A, MXF, and RFB achieved almost 3 log, >2, and

1 log kill, respectively (Figure S1), a feature that is uncommon for M. abscessus antibiotics. The magnitude of the bactericidal activity of each pair as well as the triple combination was more than the sum of their respective parts, indicative of potentiation. TBP-A + MXF showed the strongest potentiation effect, achieving more than 4 log10 CFU/ml reduction when TBP and MXF were combined at their respective MIC (Figures 1A and S1). One notable exception was the lack of RFB contribution when added to the strong TBP-MXF pair (Figure 1A and S1). Extended time-kill experiments confirmed sustained killing and absence of regrowth up to 10 days for the TBP-A + MXF and TBP-A + MXF + RFB combinations (Figure S2), suggesting the long-term efficacy potential of the regimen. The TBP-MXF pair and the triple combination were included in all subsequent analyses.

TBP-A + MXF + RFB Retains Strong Bactericidal Activity against

Reference Strains and Clinical Isolates of the M. abscessus Complex

M. abscessus presents as a complex of three subtaxa, M. abscessus subsp. abscessus (represented by the type strain M. abscessus ATCC 19977), M. abscessus subsp. bolletii, and M. abscessus subsp. massiliense. The strong bactericidal activity of TBP-A + MXF + RFB against M. abscessus subsp. abscessus was retained against representatives of the other two subspecies (M. abscessus subsp. bolletii CCUG 50184-T and M. abscessus subsp. massiliense CCUG 48898-T), as well as a panel of M. abscessus clinical isolates (Figure 1B and Table S2). At 0.5x the respective MIC of the individual drugs (Table S3), the bactericidal activity of the triple-drug combination was largely consistent across the reference strains and clinical isolates (∼2 to ∼4 log10 CFU/mL reduction; Figure 1B and Table S2).

TBP-A + MXF and TBP-A + MXF + RFB Combinations Are Strongly

Bactericidal against Nonreplicating Bacteria in Surrogate Caseum

We next asked whether TBP-A + MXF and TBP-A + MXF + RFB remain bactericidal against nonreplicating, drug-tolerant bacteria, found in caseum at the center of necrotic granulomas and cavities.^18^ To answer this question, we employed a recently developed assay in which M. abscessus is grown in a surrogate caseum matrix and adopts a nonreplicating drug-tolerant state^9,10^ (Figure S3). Nongrowing, stationary-phase cultures were then treated with the single agents and their combinations for 5 days at multiples of the respective MICs. TBP-A displayed strong bactericidal activity, reducing CFU/mL by ∼1.5 log10 at its MIC, followed by MXF. RFB displayed weak activity, consistent with previous work.^10^ The CFU reductions achieved by TBP-A alone, TBP-A + MXF, and TBP-A + MXF + RFB, at 0.5× the respective MIC of the individual drugs, were similar (Figure 1C). Thus, against nonreplicating, drug-tolerant M. abscessus in a caseum-like environment, only weak to no potentiation was observed when the drugs were combined, in contrast to the strong potentiation observed against replicating bacteria (Figure 1A,B). To put these activities in the context of concentrations expected to be achieved at the site of disease, we generated estimates of drug concentrations in caseous lung lesions at clinical doses, using a machine learning algorithm for TBP^19^ and determined experimentally for MXF^20^ and RFB.^10^ Overlaying these estimates onto the concentration kill plots (Figure 1C) indicates that TBP-A, MXF, and the featured combinations are bactericidal against caseum-induced nonreplicating M. abscessus at clinically achievable concentrations.

Lethal Secondary Effects Induced by the Single Drugs Are Retained

in the Triple-Drug Combination

Inhibition of peptidoglycan cross-linking by β-lactams and inhibition of resealing double-stranded DNA breaks by fluoroquinolones result in cell wall and DNA damage, respectively. Mycobacteria harbor the iniBAC-driven cell wall and recA-driven DNA stress response programs which are transcriptionally induced by inhibitors that cause cell wall and DNA damage, respectively.^21,22^ Cell wall and DNA damage in turn trigger repair responses, which increase the energy demand, resulting in increased metabolic activity. This drug-induced metabolic hyperdrive has been proposed to contribute to intrabacterial system collapse and subsequent cell death.^23^ β-Lactams, fluoroquinolones, and rifamycins can also induce the generation of toxic ROS, resulting in promiscuous damage of cellular macromolecules and contributing to their bactericidal activity against mycobacteria.^24−26^ Based on the strong bactericidal potentiation of the triple-drug combination, we hypothesized that these lethal secondary effects, induced by individual drugs, are retained by the TBP-A + MXF and TBP-A + MXF + RFB combinations, contributing to the strong potentiation of bactericidal activity against replicating M. abscessus.

To assess cell wall and DNA damage, we measured the fluorescence of cultures of M. abscessus ATCC 19977 reporters expressing the mCherry fluorescent protein under the control of the iniBAC or recA promoter, respectively. Cell-wall damage was confirmed by quantifying cell lysis, measuring leakage into the culture supernatant of mCherry constitutively expressed by a M. abscessus ATCC 19977 reporter. We also measured intrabacterial ATP levels as a proxy of metabolic bursts using the BacTiter-Glo assay, and intrabacterial H2O2 levels as a proxy of ROS generation using the Amplex UltraRed reagent and a M. abscessus ATCC 19977 strain expressing a variant of ascorbate peroxidase (APX). APX catalyzes the H2O2-dependent conversion of the reagent to a readily detectable fluorescent product. For each of these five readouts (detailed methods are provided as the Supporting Information), single and combination drug treatment was carried out at multiples of the respective MICs (Table S1). As expected, TBP-A and MXF induced the iniBAC and recA promoter, respectively, and the combination of TBP-A and MXF induced both promoters (Figures 2A and S4). As a transcription inhibitor, RFB interferes with inducibility of these two promoters (Figure S4) and is therefore not shown in Figure 2A. Adding MXF and/or RFB to TBP-A slightly dampened cell lysis but largely preserved TBP-A’s ability to induce profound cell-wall damage (Figure 2B and S5).

TBP-A and MXF induced a strong increase in intrabacterial ATP (∼100-fold and ∼20-fold, respectively) (Figure 2C and Figure S6). While TBP-A induced a strong (∼10-fold) increase of the intrabacterial H2O2 level, MXF and RFB increased the intrabacterial H2O2 level by only 2-fold (Figures 2D and S7), consistent with previous findings in M. tuberculosis.^25,26^ Adding RFB or MXF to TBP-A slightly dampened its strong effect on intrabacterial ATP, which was qualitatively largely preserved (Figure 2C and S6). Similarly, TBP-A alone had the strongest effect on intrabacterial H2O2 (Figures 2C,D, S6, and S7). Thus, the lethal secondary effects were not additive but were overall preserved when the drugs were combined. Collectively, these results suggest that inhibiting replication or transcription may negatively impact the generation of lethal effectors.

The impact of the ROS-scavenging antioxidant THIO on bacterial kill was measured to determine whether increased ROS generation contributes to bactericidal activity. THIO did indeed suppress the intrabacterial H2O2 increase in all treatment groups (Figures 2D and S7D) and attenuated, albeit moderately, the bactericidal activity of the single drugs and their combinations (Figures 2E and S1G,H). Interestingly, THIO exerted a pronounced growth-enhancing effect on cultures treated with the single drugs and their combinations at subinhibitory concentrations (0.25× MIC) (Figures 2E and S1G,H). Thus, increased ROS generation appears to contribute not only to the bactericidal activity but also to the growth inhibitory activity of the study drugs.

Bacteriostatic Drugs Attenuate the Bactericidal Activity and

Lethal Secondary Effects of the Bactericidal Combinations

Bacteriostatic drugs can attenuate the kill achieved by bactericidal drugs^27,28^ by dampening their drug-induced deleterious metabolic effects. The ribosome inhibitor CLR and the F-ATP synthase inhibitor BDQ are two largely bacteriostatic oral drugs that are clinically used for the treatment of M. abscessus lung disease. To determine the impact of these drugs on the bactericidal activity of TBP-A + MXF and TBP-A + MXF + RFB, time–concentration kill curves were generated with and without CLR or BDQ. CLR reduced the bactericidal activity of TBP-A + MXF and TBP-A + MXF + RFB by ∼2 log10 CFU/mL and ∼1 log10 CFU/mL, respectively, at their MIC (Figures 1A and S1CD). BDQ reduced the bactericidal activity of the combinations by ∼4 log10 CFU/mL and ∼2 log10 CFU/mL, respectively, at their MIC (Figures 1A and S1EF). We found that this negative impact of CLR and BDQ was associated with attenuation of intrabacterial metabolic burst and ROS generation (Figures 2CD, S6, and S7). BDQ exhibited a more pronounced negative effect than CLR on all lethal secondary effects (Figure 2CD), consistent with its stronger negative impact on bactericidal activity (Figure 1A). Interestingly, combining CLR or BDQ with individual drugs yielded a more complex picture (Figure 1A and Figure 2CD). Combining BDQ with TBP-A, or CLR or BDQ with MXF, decreased bactericidal activity, and these effects were generally associated with attenuation of the respective secondary lethal effects. However, combining CLR with TBP-A, or CLR or BDQ with RFB, enhanced bactericidal activity while still largely attenuating the respective lethal secondary effects. Thus, the judicious pairing of bactericidal and bacteriostatic agents may be key to the development of sterilizing drug combinations.

RFB Increases the PAE of TBP-A + MXF

PAE is the delayed regrowth of bacteria following brief antibiotic exposure, and a strong PAE may contribute to effective suppression of bacterial growth between clinical doses when plasma concentrations may fall below the MIC.^29^ Strong PAE is an attractive feature of fluoroquinolones and rifamycins in M. tuberculosis.^30,31^M. abscessus ATCC 19977 regrowth was measured over a period of 3 days, after a brief (4 h) exposure of bacterial cultures to single and combined drugs, showing PAEs of approximately 12 h for TBP-A, 4 h for MXF, and 9 h for RFB at their respective MIC (Figures 3 and S8). Adding MXF did not increase the PAE, but adding RFB to TBP-A, MXF, or the pair resulted in statistically significant increases in PAE (Figure 3). Thus, RFB exhibits two unique a long PAE and the ability to increase PAE in combination.

Triple-Drug Combination Is Required to Suppress Emergence of

Drug Resistance

A key motivation for combining antibiotics in the treatment of mycobacterial lung diseases is to slow the emergence of drug-resistant variants. To quantify the impact of combining TBP-A

• MXF and TBP-A + MXF + RFB on the frequency of resistance (FoR), we carried out spontaneous resistance selection experiments on a solid medium at 1×, 2×, 4×, or 8× the agar MIC (aMIC, lowest drug concentration that suppresses growth of colonies upon plating of 10^4^ CFU) of individual drugs (Table S1). Two independent selection experiments were carried out by plating high bacterial inocula (∼10^7^, ∼10^8^, and ∼10^9^ CFU) on agar containing the single drugs and their combinations (Table S4), amounting to ∼9 × 10^9^ CFU plated on 24 drug-containing agar plates for each drug/combination. Representative resistant colonies were restreaked to confirm resistance. The lowest concentrations that suppressed growth of wild-type colonies upon plating of high bacterial inocula were TBP-A: 2× aMIC, MXF: 4× aMIC, RFB: 2× aMIC, TBP-A + MXF, and TBP-A + MXF + RFB: combination of 2× aMIC of each drug (Table S4). The calculation of [number of resistant colonies/CFU plated on agar] at these lowest concentrations was used to infer and compare the FoR between the single drugs and their combinations. At these concentrations, TBP-A, MXF, and RFB were associated with a similar FoR of ∼10^−6^/CFU to ∼10^−7^/CFU. Unexpectedly, the TBP-A
• MXF combination exhibited a FoR of ∼10^–7^/CFU, similar to that of TBP-A and MXF alone (Table S4). Since the selection experiment did not yield any colonies resistant to the TBP-A + MXF + RFB combination, suggesting a FoR <10^–9^/CFU for the triple-drug combination, two further independent selection experiments were carried out by plating a total of ∼6 × 10^10^ CFU on 51 agar plates containing 2× aMIC of the three drugs. This yielded 6 colonies resulting in a FoR of ∼10^–10^/CFU (Table S4). Multidrug resistance of the mutants was confirmed by restreaking them on a solid medium containing each single drug and the triple combination. Thus, the TBP-A + MXF
• RFB combination effectively reduced the FoR of the single drugs by ∼1000-fold.

β-Lactam–Fluoroquinolone–Rifamycin Backbone

Is a Promising Starting Point for Optimization of Its Components

Although MXF and RFB are considered as alternative therapeutic options for treatment refractory patients, examination of their MIC distribution against M. abscessus suggests poor probability of pharmacokinetic–pharmacodynamic target attainment compared to TB.^32^ We thus hypothesized that MXF and RFB at human-equivalent doses would achieve limited efficacy in a mouse model of M. abscessus infection. Likewise, TBP-A alone is expected to demonstrate in vivo efficacy at the human equivalent dose only if combined with another β-lactam, bringing the MIC within the concentration range achieved in patients.^33^ Despite these limitations, we asked whether the 3-drug potentiation seen in vitro would translate in a mouse model. We opted for the GM-CSF^–/–^ (lacking granulocyte macrophage-colony stimulating factor) mouse model of acute M. abscessus infection because it offers robust and reproducible growth and a wide dynamic range.^34^ A drug–drug interaction study was initially carried out in CD-1 mice to ensure that exposure remained on-target when the drugs were combined. MXF 200 mg/kg and RFB 20 mg/kg were delivered at or slightly above human-equivalent exposure. TBP-A 200/200 mg/kg was selected as the efficacious dose in TB mouse models. Mice received each single drug and the triple combination starting at 2 days postinfection and for 7 consecutive days. The lung bacterial burden increased by ∼2 log10 CFU in untreated control mice between the start and end of the period. As expected, neither MXF nor RFB affected the increase in bacterial burden, whereas TBP-A prevented this increase of bacterial burden, resulting in a 2-log difference with untreated controls. Combining ineffective MXF and RFB with TBP-A resulted in increased efficacy, causing a reduction of bacterial growth by ∼3 log10 CFU compared to untreated controls. Thus, moderate potentiation was seen between TBP-A + MXF + RFB and TBP-A alone (Figure S9).

Discussion

The drug and regimen development pipeline for M. abscessus lung disease is quite thin, and incentives have been lacking, but the landscape may change in the foreseeable future, owing to fast-growing awareness and increased funding. Several safe oral drugs are in use for the treatment of M. abscessus lung disease and a few candidates are in clinical development.^4,6^ In addition, several oral repurposing candidates are emerging based on preclinical data. The preclinical drug discovery space for this neglected disease is picking up speed with several potential oral candidates lined up for clinical trials and additional candidates likely to emerge in the medium-term future. The open question is which drugs or drug classes to combine. Adding novel drugs to failing regimens will lead to the rapid emergence of resistance and loss of valuable new agents. Here, we propose a rationale and an in vitro–in vivo experimental workflow to enable the prioritization of combinations with the best sterilizing potential for clinical evaluation.

We posit that combining drugs that kill replicating and nonreplicating bacteria may overcome local or systemic immune deficiencies that are the norm in patients that contract M. abscessus infections and thus deliver sterilizing regimens.^6^ Recent work has shown that members of the β-lactam, fluoroquinolone, and rifamycin classes display bactericidal activity against both physiological states of M. abscessus.^10^ Thus, rather than aiming for incremental improvement of the suboptimal standard-of-care regimens, we selected a new regimen exclusively made of oral and safe agents, which individually display strong bactericidal activity.

Checkerboard analyses of growth inhibitory activity revealed that combining the three drugs in dual- and triple-drug combinations yielded largely additive interactions and, more importantly, no antagonistic effects. Determination of bactericidal activity against replicating reference strains of the M. abscessus complex and a panel of clinical isolates of M. abscessus revealed that dual- and triple-drug combinations potentiated the bactericidal activities of single drugs, with TBP-A + MXF and TBP-A

• MXF + RFB being the most potent. The potency of these drug combinations was retained against nonreplicating, drug-tolerant M. abscessus residing in surrogate caseum, except for RFB added to TBP-MXF at 2× MIC, showing a seemingly paradoxical effect (Figure 1C). Although RFB did not potentiate the bactericidal activity of its partner drugs in vitro, whether against replicating or nonreplicating populations, it contributed strong enhancement of their PAE, drastic reduction of the FoR by 3 orders of magnitude, and increased reduction of lung bacterial burden compared to the starting inoculum in vivo, thus demonstrating the superiority of the triple-drug combination.

Identifying drug combinations that retain or even amplify rather than attenuate the bactericidal activity of each component remains an empirical process. The effect of combining drugs on the viability of bacterial cultures needs to be measured by labor intensive, and thus throughput-limiting, CFU enumeration as the gold standard. Simple, high-throughput readouts of cell death signatures would be beneficial for the early triaging of larger numbers of potential combinations. As a first step toward the development of surrogate in vitro cell death readouts, we investigated the bacterial cell death-associated damages. The comprehensive mechanisms of action of bactericidal antibiotics involve, in addition to corruption or inhibition of the primary target, follow-on events that contribute to or are required for lethality. These responses are complex, pathogen and drug specific, and causal relationships are poorly understood, but themes are emerging. Key concepts include that bactericidal drugs often cause loss of structural integrity of the cell wall and the chromosome, and that they trigger detrimental metabolic changes such as metabolic hyperdrive and generation of toxic ROS. In addition, it has been proposed that the rate at which bacteria are killed depends on the multiplicity of targets simultaneously affected by antibiotic action and the convergence of detrimental metabolic damages downstream of inhibition of these targets.^12^

The critical assumption underlying our choice of combining a β-lactam, a fluoroquinolone, and a rifamycin is that their cell death-associated secondary effects are largely retained when they are combined (Figure 4). We confirmed the expected secondary effects of individual drugs in M. abscessus and observed that TBP is the major driver of death-associated events. We also confirmed that drug-induced ROS generation contributes to the bactericidal activity of each drug, as suggested by the attenuating effect of THIO on their bactericidal activity. Interestingly, the lethal secondary effects were retained but not amplified when drugs were combined, consistent with a model in which multiplicity of targets and convergence of secondary effects are the basis for potentiation of bactericidal activity by drug combinations (Figure 4). Our CLR and BDQ investigations suggest that bacteriostatic drugs can be detrimental to bactericidal combinations, at least in vitro, by attenuating their lethal secondary effects. However, they each presented mixed and distinct patterns of interactions, suggesting that each bacteriostatic drug should be explored separately, in vitro and in vivo.

Collectively, these results identify TBP-A + MXF

• RFB as a bactericidal backbone with attractive in vitro properties that translate into in vivo efficacy. However, none of its members was optimized against M. abscessus, as reflected by the lack of or modest efficacy of each single agent, in line with poor pharmacokinetic–pharmacodynamic profiles and probability of target attainment. This triple-drug combination template thus provides a starting point for medicinal chemistry-driven optimization. Meanwhile, TBP-A could be replaced with dual oral β-lactams, recently shown to display strong bactericidal activity against M. abscessus.^16,35,36^ MXF should be replaced with fluoroquinolones optimized against M. abscessus, and RFB could be replaced by a new, more potent RFB analogue currently in preclinical development.^37^ Furthermore, adding drugs to the combination and/or replacing single components with existing or emerging oral drugs or drug candidates from other classes can be explored.

The proposed in vitro–in vivo workflow presents several limitations and should thus be considered as an initial blueprint that requires validation. For instance, in vivo efficacy experiments were carried out only in a model of acute infection in which most bacteria actively replicate. To increase the predictive value of regimen prioritization, a model of chronic infection, where most bacteria are in a nonreplicating, stationary phase, should be included. It may also be useful to include an in vitro biofilm assay in the profiling cascade^38^ since M. abscessus can form biofilms in patient lungs.^39^

Conclusions

TBP-A + MXF + RFB is a promising combination prototype, exclusively made of oral bactericidal drugs that can be further optimized to treat M. abscessus lung disease. The cascade of assays and models employed in this work provides a useful initial framework for the preclinical profiling and prioritization of regimens for the treatment of M. abscessus lung disease. The potentiation and attenuation data can be extended to refine and further validate cell death-associated signatures of superior bactericidal activity, which could serve as early in vitro filters to enable the rational design of drug combinations with positive pharmacodynamic interactions.

Materials and Methods

Bacterial Strains, Media, and Chemicals

Bacterial strain sources are listed in Table S3. Middlebrook 7H9 (BD) supplemented with 0.2% (v/v) glycerol (Fisher Scientific), 0.05% (v/v) Tween 80 (Fisher Scientific), and 10% (v/v) Middlebrook albumin-dextrose-catalase (ADC) (BD) was used in assays unless specified otherwise. Middlebrook 7H10 (Sigma-Aldrich) supplemented with 0.2% (v/v) glycerol and 10% (v/v) Middlebrook oleic acid-albumin-dextrose-catalase (OADC) (BD) was used for growing mycobacterial strains on a solid medium. Middlebrook 7H9 and 7H10 were prepared according to the manufacturer’s instructions.

TBP, MXF, and RFB were purchased from MuseChem, Sigma-Aldrich, and Acros Organics, respectively. CLR, BDQ, avibactam, and kanamycin were purchased from MedChemExpress. THIO was purchased from Fisher Scientific. TBP, MXF, RFB, CLR, and BDQ were dissolved in 100% dimethyl sulfoxide (Fisher Scientific). Avibactam, THIO, and kanamycin were dissolved in Milli-Q water. All drugs and compounds were sterilized using Acrodisc 0.2 μm polytetrafluoroethylene membrane filters (Pall) before use.

Determination of Broth Dilution MIC of Individual Drugs in Culture

Tubes

MIC determination was carried out by using the broth macrodilution method. Briefly, 10-point twofold serial dilutions of TBP-A, MXF, RFB, CLR, or BDQ were prepared in Middlebrook 7H9 broth in 14 mL vented, round-bottom tubes (Thermo Scientific). The tubes were inoculated with the respective midexponential growing mycobacterial strain at a starting OD600 value of 0.05. The tubes were incubated at 37 °C for 3 days, with shaking on an orbital shaker, before aliquots from each tube were measured for their OD600 value using a 96-well flat, clear bottom Costar cell culture plate and an Infinite 200 Pro plate reader (Tecan). The reported broth MIC values represent the lowest concentration of the drugs that inhibited 90% of bacterial growth compared to the drug-free culture.

Checkerboard Titration Assay for Assessing Drug–Drug

Interactions in Terms of Growth Inhibitory Activity

The checkerboard titration assay for dual- and triple-drug combinations was carried out as previously described.^40,41^ Dual-drug combinations between TBP-A, MXF, and RFB were tested by adding TBP-A + MXF, TBP-A

• RFB, or MXF + RFB in Middlebrook 7H9 to a 96-well flat, clear bottom Costar cell culture plate (Corning). 2-Fold serial dilutions (0 to 25 μM) were done to allow seven different concentrations of one drug to be tested for interaction with ten different concentrations of another drug. Hence, a total of 70 different concentration combinations were tested for each pair of drugs. Combination of all three drugs was tested by adding RFB at a single concentration point to plates that were set up to test the TBP-A + MXF combination. Six different concentrations of RFB were tested (0 to 6.25 μM), resulting in 420 different concentration combinations tested for the triple-drug combination. Each 96-well plate had a culture medium-only control well and an untreated bacterial culture control well.

The prepared 96-well plates were subsequently used to determine the MIC of the individual drugs, alone and in combination, against M. abscessus subsp. abscessus ATCC 19977 via the broth microdilution method as described previously.^40^ Using the derived MIC values, the FICI was calculated to analyze the results. This calculation was done only for wells which showed 90% inhibition of bacterial culture growth compared to untreated bacterial culture wells. For pairwise drug combinations, the FICI was calculated as [MIC of drug 1 in combination/MIC of drug 1 alone]

• [MIC of drug 2 in combination/MIC of drug 2 alone]. A FICI of ≤0.5 indicates synergy, >0.5 to 4 indicates additivity (no interaction), and >4 indicates antagonism.^42^ For the triple-drug combination, the FICI was calculated as [MIC of drug 1 in combination/MIC of drug 1 alone] + [MIC of drug 2 in combination/MIC of drug 2 alone] + [MIC of drug 3 in combination/MIC of drug 3 alone]. A FICI of ≤0.75 indicates synergy, >0.75 to 4 indicates additivity (no interaction), and >4 indicates antagonism.^41^

Determination of Time–Concentration Kill Kinetic Activity

of Individual Drugs and Their Combinations against Replicating M. abscessus

Determination of time–concentration kill kinetic activity of individual drugs and their combinations was carried out as previously described, with minor modifications.^43^ Briefly, the desired concentrations of TBP-A, MXF, RFB, CLR, and BDQ were prepared in Middlebrook 7H9 broth in 14 mL vented, round-bottom tubes. The tubes were inoculated with the respective midexponential growing mycobacterial strain at a starting OD600 value of 0.005. The tubes were incubated at 37 °C for 3 days with shaking on an orbital shaker.

Determination of Bactericidal Activity of Individual Drugs and

Their Combinations against Nonreplicating M. abscessus in Surrogate Caseum

Determination of the bactericidal activity of individual drugs and their combinations against nonreplicating M. abscessus in surrogate caseum was carried out as previously described.^10^

Determination of PAE of Individual Drugs and Their Combinations

The PAE of TBP-A, MXF, RFB, and their combinations was determined as previously described with several modifications.^44^ Briefly, the desired concentrations of TBP-A, MXF, and RFB were prepared in Middlebrook 7H9 broth in 14 mL vented, round-bottom tubes. The tubes were inoculated with midexponential growing M. abscessus subsp. abscessus ATCC 19977 at a starting OD600 value of 0.2. The tubes were then incubated at 37 °C for 4 h. Subsequently, the cultures were then washed two times via centrifugation and resuspension of the cell pellets in Middlebrook 7H9. The tubes were then incubated at 37 °C for 3 days, with shaking on an orbital shaker. At selected time points, aliquots from each tube were measured for OD600 using a 96-well flat, clear bottom, Costar cell culture plate and the Infinite 200 Pro plate reader. The duration of the PAE was calculated as the time taken for the drug-treated culture to reach 50% of the maximum OD600 of the drug-free culture minus the time taken for the drug-free control to reach the same point.^45^

Assays to Measure Cell Wall and DNA Damage, Cell Lysis, Intrabacterial

ATP, and H2O2 Levels

All assays and methods are provided in the Supporting Information.

Determination of aMIC

The aMIC of TBP-A, MXF, and RFB was determined by the agar dilution method according to the CLSI protocol,^46^ as described in Table S1.

Selection of Spontaneous Resistant Strains against Individual

Drugs and Their Combinations

Spontaneous resistant strains were selected against TBP-A, MXF, RFB, and the TBP-A + MXF and TBP-A

• MXF + RFB combinations as described previously.^43^ For selection against the individual drugs and the TBP-A
• MXF combination, ∼10^7^, ∼10^8^, and ∼10^9^ CFU of exponentially growing M. abscessus subsp. abscessus ATCC 19977 were plated on Middlebrook 7H10 agar plates containing 1×, 2×, 4×, or 8× aMIC of TBP-A, MXF, or RFB (Table S1), either as individual drugs or in drug combinations (Table S4). This resulted in a total of ∼9 × 10^9^ CFU plated on 24 drug-containing agar plates for each drug/combination selection. For selection of strains resistant to the TBP-A + MXF + RFB combination, a total of ∼6 × 10^10^ CFU of exponentially growing M. abscessus subsp. abscessus ATCC 19977 was plated on 51 Middlebrook 7H10 agar plates containing 2× aMIC of TBP-A, MXF, and RFB in combination (Table S4; refer to the Results section “Triple-Drug Combination Is Required to Suppress Emergence of Drug Resistance” for the rationale behind this selection condition). The agar plates were incubated for 7 days at 37 °C. Representative resistant colonies that grew were re-streaked on agar containing the same drug(s) and concentration(s) at which they were isolated to confirm resistance.