The cytochrome bd-type quinol oxidase is important for survival of Mycobacterium smegmatis under peroxide and antibiotic-induced stress

Targeting respiration and ATP synthesis has received strong interest as a new strategy for combatting drug-resistant Mycobacterium tuberculosis. Mycobacteria employ a respiratory chain terminating with two branches. One of the branches includes a cytochrome bc1 complex and an aa3-type cytochrome c oxidase while the other branch terminates with a cytochrome bd-type quinol oxidase. In this communication we show that genetic inactivation of cytochrome bd, but not of cytochrome bc1, enhances the susceptibility of Mycobacterium smegmatis to hydrogen peroxide and antibiotic-induced stress. The type-II NADH dehydrogenase effector clofazimine and the ATP synthase inhibitor bedaquiline were bacteriostatic against wild-type M. smegmatis, but strongly bactericidal against a cytochrome bd mutant. We also demonstrated that the quinone-analog aurachin D inhibited mycobacterial cytochrome bd at sub-micromolar concentrations. Our results identify cytochrome bd as a key survival factor in M. smegmatis during antibiotic stress. Targeting the cytochrome bd respiratory branch therefore appears to be a promising strategy that may enhance the bactericidal activity of existing tuberculosis drugs.

The respiratory chain enzyme complexes that are part of the oxidative phosphorylation pathway establish a proton motive force across the bacterial cytoplasmic membrane and ATP synthase utilizes the energy of the proton motive force for synthesis of ATP. Mycobacterial ATP synthase has been validated as target of bedaquiline (BDQ), the lead compound of the diarylquinoline class of drugs, which selectively inhibits this enzyme in a variety of mycobacterial strains [12][13][14][15][16] . BDQ has received accelerated approval by the US Food & Drug Administration (FDA) and the European Medicines Agency (EMA) for treatment of MDR-TB 17,18 . Moreover, components of the respiratory chain such as the type-II NADH dehydrogenase (NDH-2) and the cytochrome bc 1 complex are targeted by small-molecule compounds that are currently in clinical development [19][20][21][22][23][24] . Mycobacteria have a branched electron transport chain. Electrons from the menaquinone pool can be passed on either to the cytochrome bc 1 complex, which forms a supercomplex with the cytochrome aa 3 oxidase, or alternatively to the cytochrome bd-type quinol oxidase 9,10,20 (Fig. 1). Both branches transfer the electrons onto molecular oxygen, yielding H 2 O, but they differ in the efficiency of energy conservation. The cytochrome bc 1 /aa 3 branch establishes a higher proton motive force as compared with the cytochrome bd branch and consequently is energetically more efficient. Therefore, this respiratory branch is mainly utilized during aerobic, replicating conditions 25,26 .
Genetic knock-out of the cytochrome bc 1 complex in M. smegmatis substantially decreased the growth rate of the bacteria under aerobic growth conditions, while knock-out of cytochrome bd did not 25,27 . The cytochrome bc 1 complex has also been validated as target of the imidazopyridine class of drugs 22,24 , whereas no antibacterials targeting cytochrome bd have been reported yet. These findings point towards cytochrome bc 1 /aa 3 as the more promising drug target of the two respiratory chain branches in mycobacteria. However, the proteins of the cytochrome bc 1 /aa 3 branch are down regulated during hypoxia and chronic infection in a mouse model, while these conditions induced the expression of cytochrome bd, suggesting an important role for this enzyme in respiration during hypoxia 26 . Additionally, cytochrome bd was induced when the cytochrome bc 1 complex was impaired due to deletion mutations 25 , upon inhibition by small-molecule drugs 28 or when cytochrome c maturation was disturbed 29 , suggesting that the cytochrome bd branch may (partially) be able to compensate for lack of function of the cytochrome bc 1 /aa 3 branch of the respiratory chain 25,28,29 .
Cytochrome bd can also play a role in protection against different types of stress [30][31][32] . In Escherichia coli, exposure to exogenous hydrogen peroxide and nitric oxide induced expression of cytochrome bd and strains lacking cytochrome bd were found hyper-sensitive to peroxide and nitrosative stress [33][34][35][36] as well as to low iron concentrations 37 . In M. tuberculosis, cytochrome bd expression in the mouse lung is upregulated during chronic infection 26 . During an inflammatory reaction, macrophages in the host can produce reactive oxygen species (ROS) to kill engulfed bacteria. Overexpression of cytochrome bd in M. tuberculosis is associated with increased peroxide resistance 29 . Upregulation of cytochrome bd may represent a protection mechanism to survive the host's immune response. These data point towards cytochrome bd as an important contributor to stress resistance in (myco-) bacteria.
In this study, the role of the two mycobacterial respiratory chain branches in response to stress elicited by peroxides and antimicrobials was investigated. For this aim we challenged strains of M. smegmatis lacking cytochrome bd or the cytochrome bc 1 complex in vitro with these stress factors to elucidate the importance of each respiratory chain branch in protection against them.

Results
Bioenergetic parameters of Mycobacterium smegmatis strains with inactivated respiratory chain branches. The role of two respiratory chain branches in mycobacteria was investigated using mutant strains impaired in one of the two branches. These strains maintain either only the cytochrome bd branch (strain Δ qcrCAB :: hyg) or the cytochrome bc 1 /aa 3 branch (strain Δ cydA :: kan) (Fig. 1). The growth rate of the Δ cydA :: kan strain was comparable to that of the wild-type, whereas growth of the Δ qcrCAB :: hyg strain was substantially lower ( Fig. 2A), confirming previous data 25,27 . We then extended the earlier reported microbiological characterization of the mutant strains and determined central bioenergetic parameters for the two mutants. Cellular ATP levels were unaltered in the Δ cydA :: kan mutant as compared with the wild-type, but were decreased by ~40% in the Δ qcrCAB :: hyg strain (Fig. 2B). Similarly, oxygen consumption rates in inverted membrane vesicles isolated form aerobically grown cells were almost unchanged in the Δ cydA :: kan mutant versus wild-type, but lower in Δ qcrCAB :: hyg (Fig. 2C). These results reflect the higher respiratory efficiency of the cytochrome bc 1 /aa 3 branch. Based on growth rate and bioenergetic characterization the cytochrome bc 1 /aa 3 branch can be regarded as the more promising target pathway of the two branches.
Sensitivity for hydrogen peroxide stress. Next, we investigated the importance of the two respiratory chain branches in response to peroxide stress. Exponentially growing M. smegmatis cells were exposed to hydrogen peroxide (20 mM, final conc.) for various time intervals and colony-forming units were enumerated. Incubation with hydrogen peroxide had a bacteriostatic effect on wild-type M. smegmatis and for the Δ qcrCAB :: hyg mutant a minor decrease in viability was found (Fig. 3). For the Δ cydA :: kan mutant, a 99% decline in cell viability was observed after 60 min exposure (Fig. 3). These results suggest that cytochrome bd plays a protective role during oxidative stress in M. smegmatis, whereas the cytochrome bc 1 complex is of minor importance for survival under these conditions. Sensitivity for the NDH-2 effector clofazimine. We hypothesized that mycobacteria with impaired respiratory chain branches may also be more sensitive to antimicrobials that cause production of reactive oxygen species. Clofazimine (CFZ) is a front-line anti-leprosy drug that presently is repurposed for usage against tuberculosis. CFZ is an electron carrier that interferes with the type II NADH dehydrogenase (NDH-2) in mycobacteria 19 . As such, it can transfer electrons from NDH-2 directly to oxygen, thereby producing ROS 19 . First, we confirmed that CFZ caused time-dependent development of ROS by inverted membrane vesicles from the M. smegmatis wild-type strain used in our laboratory (Supplementary Figure  S1). Subsequently we investigated if either cytochrome bd or the cytochrome bc 1 complex can protect M. smegmatis against CFZ. For this purpose the bacteria were incubated for 72 hours in liquid culture with varying concentrations of the drug. CFZ was bacteriostatic against the wild-type strain, even at the highest concentration investigated (25x MIC, 7.5 μ g/mL) (Fig. 4). The Δ qcrCAB :: hyg mutant showed marginally higher sensitivity for CFZ as compared with the wild-type (Fig. 4). However, the viability  of the Δ cydA :: kan mutant was strongly reduced in response to CFZ challenge. CFZ at concentrations > 0.3 μ g/mL was bacteriostatic for the Δ cydA :: kan mutant and concentrations > 1.5 μ g/mL were bactericidal. With 7.5 μ g/mL CFZ the limit of detection was reached after 72 hours of exposure (Fig. 4). These results indicate that cytochrome bd, but not the cytochrome bc 1 complex, can protect the bacteria against the bactericidal effect of clofazimine. We hypothesized that the increased sensitivity of the Δ cydA :: kan strain was due to ROS production by CFZ. To test this hypothesis we investigated the effect of chlorpromazine (CPZ), a phenothiazine-class drug that inhibits type-II NADH dehydrogenase 20,23 , but does not produce ROS 19 , on wild-type and the Δ cydA :: kan mutant. As expected, CPZ did not discriminate between wild-type M. smegmatis and the Δ cydA :: kan mutant (Supplementary Figure S1). Sensitivity for the ATP synthase inhibitor bedaquiline. The results described above demonstrate that genetic inactivation of cytochrome bd, but not of the cytochrome bc 1 complex, converts the bacteriostatic effect of hydrogen peroxide and of clofazimine into a bactericidal effect. Next, we expanded our experiments to the ATP synthase inhibitor bedaquiline (BDQ). Whereas BDQ is bactericidal against M. tuberculosis, it is bacteriostatic against M. smegmatis 12 . A transcriptional and proteomic analysis recently revealed that treatment of M. tuberculosis with BDQ triggers strong upregulation of cytochrome bd 38 and deletion of cytochrome bd in M. tuberculosis enhanced the bactericidal activity of BDQ 39 . We therefore investigated if genetic inactivation of one of the respiratory chain branches would convert the bacteriostatic activity of BDQ on M. smegmatis into bactericidal activity.
BDQ was bacteriostatic against wild-type M. smegmatis, even at the highest concentration used (300x MIC, 5 μ g/mL) (Fig. 5). The Δ qcrCAB :: hyg strain was less sensitive to BDQ as compared with the wild-type strain (Fig. 5). However, in case of the Δ cydA :: kan mutant, challenge with BDQ (1 μ g/mL) led to a ~1 log 10 reduction in colony forming units and 5 μ g/mL BDQ caused ~3 log 10 kill, approaching the limit of detection after 3 days of treatment (Fig. 5). Cytochrome bd thus protects M. smegmatis against killing by bedaquiline, whereas the cytochrome bc 1 /aa 3 branch does not. We attempted to link the protective function of cytochrome bd to production of ROS in the presence of BDQ, however, inverted membrane vesicles from M. smegmatis did not show increased ROS formation after treatment with BDQ (Supplementary Figure S1).
The results obtained for CFZ and BDQ demonstrate that inactivation of the cytochrome bd branch, but not of the cytochrome bc 1 /aa 3 branch, can convert bacteriostatic activity of an antibacterial drug into bactericidal activity. Our findings identify cytochrome bd as an important survival factor in mycobacterial metabolism.
Inactivation of mycobacterial cytochrome bd by a small-molecule inhibitor. Genetic inactivation of cytochrome bd can considerably increase the potency of two prominent antibacterial drugs, CFZ and BDQ. Based on these findings we tested if small-molecule inhibitors can block the activity of cytochrome bd in M. smegmatis. The aurachin class of quinone analogs has been reported as inhibitors of a variety of quinone-modifying enzyme [40][41][42] . Within this class, aurachin D was previously shown to preferentially inhibit E. coli cytochrome bd as compared with other quinone-modifying enzymes 42 . We investigated the effect of aurachin D on the oxygen consumption activity of inverted membrane vesicles from M. smegmatis. Aurachin D inhibited oxygen consumption in a dose-dependent manner with 50% maximal inhibition for wild-type strain (Fig. 6). Interestingly, this inhibitory effect was clearly stronger in membrane vesicles of the Δ qcrCAB :: hyg strain, where ~90% maximal inhibition was reached (IC 50 ~400 nM) (Fig. 6). This suggests that the main target in mycobacterial oxidative phosphorylation was cytochrome bd.
Subsequently, we evaluated the effect of aurachin D on mycobacterial growth. We found that for all three strains tested (wild-type, Δ cydA :: kan, Δ qcrCAB :: hyg) the minimal inhibitory concentrations (MICs) were > 85 μ M (data not shown). This result suggests that the inhibitor is not capable of effectively crossing the mycobacterial cell envelope.

Discussion
Previously it has been reported that genetic inactivation of cytochrome bd considerably decreased virulence or survival in the host of a variety of pathogenic bacterial strains. In Shigella flexneri, Brucella abortus and Salmonella enterica Serovar Typhymurium, the causative agents of bacterial dysentery, brucellosis and typhoid fever, inactivation of cytochrome bd considerably impaired intracellular survival and virulence [43][44][45] . In Klebsiella pneumonia cytochrome bd was found crucial for free energy transduction under microaerobic conditions and for protection of anaerobic processes such as nitrogen fixation 46 . In case of group B streptococci, inactivation of cytochrome bd led to decreased growth in human blood 47 . Cytochrome bd may also allow strictly anaerobic bacteria such as Bacteriodes fragilis to survive under nanomolar oxygen concentrations, potentially facilitating survival of opportunistic pathogens in the host 48 .
In this study, we evaluated the function of the two mycobacterial respiratory chain branches in response to stress. The cytochrome bc 1 complex is a validated drug target in M. tuberculosis 22,24 , however, upregulation of cytochrome bd may partially compensate for inhibition of cytochrome bc 1 function 25,28,29 . Therefore, it has been postulated that simultaneously targeting both respiratory chain branches with inhibitors might be required to effectively disrupt mycobacterial respiration 29 . Whereas the cytochrome bd branch may in part be able to compensate for inactivation of the cytochrome bc 1 complex, our results indicate that the cytochrome bc 1 /aa 3 branch is not able to compensate for loss of cytochrome bd functionality. Inactivation of cytochrome bd, although not directly leading to a phenotype, exerts a strong impact on bacterial viability in the presence of antibiotic stress. This highlights the importance of the cytochrome bd branch as a survival factor in M. smegmatis and suggests that targeting this terminal oxidase may be a successful strategy for weakening the mycobacterial stress response.
The hypersensitivity of the cydAB mutants to exogenous hydrogen peroxide is not due to impaired growth of the mutant strain, since growth rate and ATP levels are similar to the wild-type. Giuffre, Borisov and colleagues suggested two molecular mechanisms for peroxide protection by cytochrome bd in E.coli 32 . First, cytochrome bd as oxygen scavenger may decrease the intracellular oxygen tension, thereby preventing the formation of reactive oxygen species. Second, cytochrome bd displays catalase activity 32,34 and might thus directly metabolize peroxides. Both mechanisms may contribute to the protective role of cytochrome bd against hydrogen peroxide stress in M. smegmatis and their respective importance in mycobacteria needs to be further elucidated.
Our experiments revealed that cytochrome bd plays an important role in protection against two prominent anti-tuberculosis drugs, both targeting oxidative phosphorylation. Protection against clofazimine, a ROS-producing drug, is most likely due to the ability of cytochrome bd to metabolize and/or prevent formation of peroxides. Our data do not allow for pinpointing the mechanism of protection against BDQ. Inhibition of ATP synthase may well result in reduction of the electron flow through the respiratory system. As a result, the reduction state of the respiratory complexes increases which in turn leads to increased production of ROS. Higher cellular NADH/NAD + ratios and enhanced expression of bacterioferritin, indicating BDQ-induced backpressure and ROS formation, have previously been reported for M. tuberculosis treated with BDQ 38 . However, it is possible that the levels of ROS produced by BDQ are not high enough for detection in case the membrane vesicles used in our study are leaky. Alternatively, protection by cytochrome bd may be due to its lack of proton pump functionality. Cytochrome bd in E. coli has been found electrogenic, but displays a low H + /eratio 49,50 . In this way cytochrome bd may alleviate membrane hyperpolarization.
Inactivation of cytochrome bd converts the bacteriostatic activity of clofazimine and bedaquiline against M. smegmatis into strong bactericidal activity. This finding may be of pharmaceutical and clinical relevance as the bacteriostatic activity of bedaquiline is not restricted to M. smegmatis, but also found for pathogenic non-tuberculous mycobacterial strains, such as the M. avium complex 51 . These pathogenic strains typically show only low susceptibility towards current antibacterial chemotherapy 52 . Inactivation of cytochrome bd may assist in improving treatment options for infections caused by these recalcitrant bacteria. It would be important to assess if cytochrome bd deletion mutants in these pathogenic bacteria display increased sensitivity to (ROS-producing) antibacterials as well.
Inhibition of mycobacterial cytochrome bd by aurachin D serves as proof-of-concept for small-molecule inhibition of this important new drug target. Improved aurachin derivatives with better ability to penetrate the mycobacterial cell envelope may constitute a new class of anti-tubercular drugs. Cytochrome bd is of particular interest as potential drug target, as it is only found in prokaryotes. The absence of a human homologue may facilitate selective targeting. However, whole-cell screening on bacteria under aerobic, replicating conditions, which typically are applied for high-throughput discovery procedures 53 , may not allow for detection of cytochrome bd inhibition. Screening for bacteria under stressed conditions, e.g. in the presence of hydrogen peroxide or bedaquiline, may be applied as an alternative. Additionally, target-or pathway-based screenings, e.g. based on the inverted membrane vesicle system described in this report, against chemical libraries might lead to the discovery of potent cytochrome bd inhibitors. Growth curves. Growth curves for wild-type and mutant M. smegmatis were determined using a 96-well plate system. Bacteria were diluted to an optical density at 600 nm of 0.01 and optical density was determined at 20 minute intervals for 60 hours. The optical density was measured with a UV-VIS spectrophotometer (Varian Cary50).

Preparation of inverted membrane vesicles.
Inverted membrane vesicles (IMVs) of the bacterial strains were prepared as described previously 54 . Briefly, M. smegmatis was grown for three days in a pre-culture to late-exponential phase. Cells were sedimented by centrifugation at 6000 x g for 20 minutes. The pellet was washed with phosphate buffered saline (PBS, pH 7.4) and centrifuged at 6000 x g for 20 min. Each 5 g of cells (wet weight) was re-suspended in 10 mL of ice-cold lysis buffer (10 mM HEPES, 5 mM MgCl 2 and 10% glycerol at pH 7.5) including protease inhibitors (complete, EDTA-free; protease inhibitor cocktail tablets from Roche). Lysozyme (1.2 mg/mL), deoxyribonuclease I (1500 U, Invitrogen) and MgCl 2 (12 mM) were added and cells were incubated with shaking for one hour at 37 °C. The lysates were passed three times through a One Shot Cell Disruptor (Thermo Electron, 40 K) at 0.83 kb to break Scientific RepoRts | 5:10333 | DOi: 10.1038/srep10333 up the cells. Unbroken cells were removed by three centrifugation steps (6000 x g for 20 min at 4 °C). The membranes were pelleted by ultracentrifugation at 222,000 x g for one hour at 4 °C. The pellet was re-suspended in lysis buffer and snap-frozen until use. The protein concentration was measured using the BCA Protein Assay kit (Pierce) as described by the manufacturer.

Oxygen respiration assays. Oxygen respiration and the effect of inhibitors on oxygen respiration
were measured by polarography using a Clark-type electrode. The electrode was fully aerated (212 μ M O 2 at 37 °C) and calibrated with sodium hydrosulfite. The inverted membrane vesicles were pre-incubated for three minutes with the inhibitors in a pre-warmed (37 °C) buffer containing 50 mM MES and 2 mM MgCl 2 (pH 6.5). NADH was added as electron donor to a final concentration of 250 μ M and oxygen respiration was measured for 90 seconds. Potassium cyanide was used as a control for inhibition. Two independent experiments were performed and average values plus standard errors were calculated.
Cellular ATP levels were determined using the luciferase bioluminescence method described previously 55 . Briefly, 1.0-mL samples taken from M. smegmatis cultures grown as described above were centrifuged at 8000 * g for 10 min. The pellets were re-suspended in 50 μ l water and a 10-fold volume of boiling 100 mM TRIS-HCl, 4 mM EDTA (pH 7.75) was added. After incubation at 100 °C for 2 min the samples were centrifuged (1000 * g, 60 s) and the supernatants transferred to fresh tubes. 100 μ l luciferase reagent (ATP Bioluminescence assay, Roche) was added to 100 μ l sample and luminescence was measured with a Luminometer (LKB).
Hydrogen peroxide and antibiotic sensitivity assays. Bacterial strains were grown to an optical density at 600 nm of 0.5. For hydrogen peroxide sensitivity assays, hydrogen peroxide (30% (w/v) stock) was added to an Eppendorf tube containing 0.49 mL of bacterial suspension to a final concentration of 20 mM. After the indicated time of incubation at 37 °C with shaking, 15 μ l of catalase (10 mg/mL) was added to degrade hydrogen peroxide and thereby stop the reaction. For antibiotic sensitivity assays, 10 mL of bacterial cultures were incubated with the antibiotic for three (clofazimine and chlorpromazine) or four days (bedaquiline) at 37 °C with shaking. All samples were diluted in PBS and 0.1 mL was plated on 7H10 agar plates, containing oleic acid (0.05 g/l) and 10% Middlebrook albumin dextrose catalase enrichment (BBL). Cell viability was measured by counting colony-forming units per mL (CFU/mL) after 72 h (wild-type and Δ cydA :: kan strain) or 96 h (Δ qcrCAB :: hyg strain) incubation at 37 °C. The limit of detection was 100 CFU/mL. Survival was determined as percentage of surviving cells compared to untreated cells at day 0.

ROS detection assays.
For detection of reactive oxygen species the Amplex Red® Hydrogen Peroxide/ Peroxidase Assay kit (Invitrogen) was used as described by the manufacturer with minor modifications. To measure ROS production in inverted membrane vesicles, 1 mL samples of 0.05 M sodium phosphate, pH 7.4 containing 20 μ g M. smegmatis inverted membrane vesicles, 0.2 mM NADH, 50 μ M Amplex Red®, 2 U horseradish peroxidase (HRP), 80 U superoxide dismutase (SOD) and the antibiotic diluted in DMSO in 1x reaction buffer (0.05 M sodium phosphate, pH 7.4) were prepared. Superoxide dismutase was added to allow for detection of superoxide. ROS production was determined by measuring absorbance at 563 nm for 30 minutes with a UV-VIS spectrophotometer (Varian Cary50).