Introduction

Tuberculosis (TB) is a global health threat further worsened by increasing incidences of drug resistance and HIV co-infection1. New drugs to combat resistant TB and to shorten the current 6 to 9 month standard-of-care chemotherapy are urgently needed. In most cases, infection starts in the lungs where Mycobacterium tuberculosis induces well-structured granulomas composed of various hematopoietic cells in which the pathogen is contained2, 3. During progression to active disease, infected macrophages and neutrophils undergo necrosis at the core of these structure, resulting in the formation of a hypoxic caseum2.18F-fluorodeoxyglucose positron emission tomography/computed tomography has revealed that pulmonary TB is a highly dynamic disease: coexisting inflammatory hotspots in lungs of TB patients increase in size, shrink or even resolve over time while others are newly formed4, 5. Recent studies have indicated that pro- and anti-inflammatory signals are spatially segregated within granulomas of humans suggesting that granuloma formation in TB is a direct consequence of the local activity of inflammatory pathways6.

Among the available animal models, non-human primates mirror the facets of human TB best, but broad application is hampered by prohibitive cost and ethical concerns. To date, mice remain the most widely used model for pre-clinical TB drug development. However, upon M. tuberculosis infection, small rodents develop a more uniform type of unstructured lesions that lack central necrosis and hypoxia7, two key features of human granulomas. Commonly employed strains such as BALB/c mice may therefore inaccurately predict the outcome of new drug candidates and treatment regimens in humans, as emphasized by the disappointing results of three large and costly clinical trials8,9,10. Bridging the disconnect between pre-clinical mouse models and TB patients is therefore mandatory for accelerated drug development.

The murine supersusceptibility to tuberculosis 1 (sst1) locus has been found to prevent formation of necrotic lesions in mice. The intracellular pathogen resistance 1 (Ipr1)-protein encoded within this locus can switch the molecular program of infected macrophages from necrosis to apoptosis11. C3HeB/FeJ mice lacking a functional ipr1 gene develop large pulmonary lung lesions with central caseous necrosis during M. tuberculosis infection12. This model has been used in several studies to evaluate monotherapy with anti-TB drugs and clinically relevant drug regimens12,13,14,15,16,17. In-depth pathological characterization of M. tuberculosis-infected C3HeB/FeJ mice revealed multiple lung lesion types which are either large, caseous and necrotic and characterized by neutrophil-dominated granulocytic pneumonia or which reflect unstructured lesions as observed in BALB/c mice18. C3HeB/FeJ mice with a predominance of caseous necrotic lesions versus cellular inflammatory lesions (‘BALB/c-type’) respond differently to chemotherapy14.

TB drugs need to penetrate the granuloma to act on the pathogen in different physiological states. Previous in vitro studies suggest that drug penetration into non-replicating M. tuberculosis is reduced19. Such non-replicating persisters are a difficult-to-treat population believed to reside in the unfavourable microenvironment of necrotic and hypoxic granulomas2. Moreover, drugs have been shown to differentially penetrate into well-structured TB lesions in rabbits20 and in humans21. Moxifloxacin (MXF) in particular was demonstrated to diffuse relatively poorly into central caseum – a finding which can explain the failure of this drug to shorten treatment duration in the clinic9, 10, 21. Thus, a substantial body of evidence suggests that different granuloma pathologies as they occur in TB patients impact treatment outcomes.

The inducible oxygen-dependent host enzyme nitric oxide synthase-2 (NOS2) counteracts microbial intruders by means of NO production. Consequently, Nos2 −/− mice deficient for this defence mechanism are exquisitely susceptible to bacterial infections including TB22. Our previous work demonstrated that M. tuberculosis-infected Nos2 −/− mice develop solid structured granulomas in which central necrosis followed by hypoxia can be induced by neutralization of interferon-gamma or tumour necrosis factor alpha (TNF-α)23, 24. The human-like granuloma pathology of Nos2 −/− mice prompted us to harness this model for evaluation of TB monotherapy with first-line drugs and new candidates under development, to examine the direct impact of necrosis and hypoxia on outcome of chemotherapy.

Results

Dermal M. tuberculosis infection of Nos2 −/− mice consistently induces lung lesions with central necrosis and hypoxia

Previously, we demonstrated the capacity of Nos2 −/− mice to develop a TB patient-like pulmonary pathology upon dermal M. tuberculosis infection that includes central necrosis and hypoxia in structured lesions23, 24. In order to harness Nos2−/− mice for TB drug development as schematically depicted in Fig. 1a, the temporal onset of patient-like pathology needed to be defined. We dermally infected groups of mice, neutralized the TNF-α response at 14 and 21 days post-infection, and assessed bacterial burden (superior, middle, inferior and post-caval lobe) and pathology (left lobe) of lungs at days 42, 56, 70 and 84. The total number of bacilli residing in lungs of Nos2−/− mice was 105 colony forming units (c.f.u.) at day 42 and increased to 107 c.f.u. by day 70 after which the bacterial burden remained stable on a level comparable to low-dose-infected BALB/c mice (Fig. 1b)25. Consecutive thin sections of the fixed and paraffin-embedded left lung lobe were stained with H&E or used for immunohistochemistry to visualize hypoxic areas. The average number of lesions per section moderately increased from 6.3 at day 42 to 8.5 at the end of the experiment (Fig. 1c). In contrast, the numbers of necrotic and hypoxic lesions were negligible 42 days post-infection, became apparent at day 56 and further increased until 84 days, when roughly 75% of all lesions showed patient-like pathology including necrosis and hypoxia (Fig. 1d,e). Importantly, lung lesions of all mice collectively and co-ordinately developed advanced pathology in a distinctively narrow time window. We concluded that the response of hypoxic necrotizing pulmonary lesions to TB chemotherapy in Nos2−/− mice could be assessed by initiating drug treatment at day 42 in the control group and after onset of necrosis and hypoxia at day 56 in the study group (Fig. 1a).

Figure 1
figure 1

Temporal development of hypoxic necrotizing lung lesions in Nos2 −/− mice upon dermal M. tuberculosis infection. (a) Schematic representation of study design. (b) Infection dose (day 0) and bacterial burden in lungs at 42, 56, 70 and 84 days post-intradermal M. tuberculosis infection determined by plating serial dilutions of lung homogenates on agar. n = 10, representative of 2 independent experiments. (c–e) Histopathological evaluation of sections obtained from the entire left lung lobe of M. tuberculosis-infected Nos2 −/− mice at designated time points. Consecutive sections were stained with H&E to quantify lesions (c) and necrosis (d). Hypoxic lesions were identified by detection of pimonidazole adducts (Supplementary Fig. S1e), n = 5, representative of 3 independent experiments. Data are represented as means ± s.d. TNF-α, tumour necrosis factor alpha.

Full size image

Pulmonary pathology of M. tuberculosis-infected Nos2 −/− mice includes central necrosis, hypoxia, foamy macrophages and peripheral fibrosis

With few exceptions, lung lesions 42 days after infection appeared non-necrotizing and not hypoxic (Supplementary Fig. S1a). At day 56, about 50% of all lesions showed signs of central necrosis of which half were hypoxic (Supplementary Fig. S1a,b). Pimonidazole adducts formed under oxygen limitation often appeared in regions with circular shape. (Supplementary Fig. S1a, lower row, Supplementary Fig. S1b, middle panel). Acid-fast staining of lung sections revealed a large number of M. tuberculosis bacilli in lesion centres at onset of necrosis (Supplementary Fig. S1b, right panel). From day 70 onwards, lighter stained central areas of lung lesions were enlarged and sharply separated from the surrounding granulomatous inflammatory infiltrates indicating advanced necrosis (Supplementary Fig. S1a, upper panel).

On day 84 we observed advanced necrosis characterized by enlarged areas of central tissue consolidation, fibrous encapsulation and caseation of lesions on day 84, which are pathological late-stage correlates of human TB (Fig. 2a,b). The fibrous capsule was separated from the central caseum by a layer of infected cells presumably foamy macrophages (Fig. 2c,d) as seen in human granulomas26. In contrast to other murine TB models, lung lesions of Nos2−/− mice collectively and coordinately developed hypoxic necrotizing lung lesions allowing to assess the direct impact of TB chemotherapy on patient-like pathology by initiating drug treatment at different time points (Fig. 1a).

Figure 2
figure 2

Nos2 −/− mice with TB form pulmonary caseous lesions with fibrous encapsulation and infected cells. (a–d) Lung sections of Nos2 −/− mice 84 days post-intradermal M. tuberculosis infection. H&E, Trichrome and Ellis’ acid fast staining of caseous lesions are shown. (a) The caseous center is separated from the fibrous cuff consisting of blue-stained newly formed collagen fibers (b) by a layer of lighter-appearing cells (c) presumably including infected foamy macrophages. (d) Acid fast rod-shaped M. tuberculosis bacilli are stained pink (arrows). Abbreviations: C, caseation; I, intersection; FC, fibrous cuff; H&E, haematoxylin & eosin.

Full size image

Efficacy of isoniazid monotherapy is reduced in hypoxic necrotizing lesions

The first-line TB drug isoniazid (INH) targets mycobacterial cell wall biosynthesis and kills growing M. tuberculosis but is well tolerated by the non-replicating pathogen in vitro 27, 28 and in vivo 29. INH at a dose of 25 mg/kg was strongly bactericidal in Nos2 −/− mice when chemotherapy was started 42 days post-infection (Fig. 3a). The initial steep decline in pulmonary bacterial burden slowed down after 14 days of drug treatment revealing a bi-phasic profile comparable to observations made in BALB/c mice (Fig. 3a)13. In contrast, when INH administration in Nos2 −/− mice was initiated after onset of necrosis and hypoxia the bacterial load declined by 1.62 log c.f.u. during the first 2 weeks of treatment and remained an a comparable level thereafter (Fig. 3a). Development of genetic drug resistance that might have prevented further decline in c.f.u. could be ruled out since plating lung homogenates on agar containing 4 μg/ml INH at the end of the experiment did not result in colony formation. We conclude that a drug-tolerant persister population of M. tuberculosis developed in hypoxic necrotizing lung lesions of Nos2 −/− mice under INH pressure.

Figure 3
figure 3

Impact of monotherapy on pulmonary bacterial burden in Nos2 −/− mice with TB. Monotherapy for 4 weeks was initiated at day 42 (control group) or after onset of central necrosis and hypoxia in lung lesions at day 56 (study group) as described in Fig. 1a. Drugs were administered by oral gavage on 6 days per week. Bacterial burdens in lungs were assessed by plating organ homogenates on agar. One representative dataset out of 2 independent experiments is shown. Data were analysed using two-way ANOVA with multicomparison and Tukey’s post-test. Shown are means ± s.e.m. and the log10 reduction of mean c.f.u. counts as compared to animals prior to chemotherapy (summarized in Supplementary Table S1), n = 5, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. (a) isoniazid, 25 mg/kg. (b) rifampin, 10 mg/kg and 20 mg/kg; rifapentine, 5 mg/kg and 10 mg/kg. (c) metronidazole, 400 mg/kg. (d) pretomanid, 75 mg/kg. (e) delamanid, 1 mg/kg. (f) linezolid, 100 mg/kg. (g) moxifloxacin, 200 mg/kg. (h) BTZ043, 50 mg/kg.

Full size image

Comparable bactericidal activity of rifampin and rifapentine

Rifamycins inhibit RNA biosynthesis and are a cornerstone of current TB chemotherapy30. While rifampin (RIF) has been in clinical use for decades, its derivative rifapentine (RPT) showed higher potency and greater drug exposure in mice, thereby raising hopes for shorter treatment over RIF31,32,33. Both drugs, RIF and RPT were profoundly bactericidal in Nos2 −/− mice (Fig. 3b). We observed a significant dose response for RIF administered at 10 or 20 mg/kg, which was only moderately influenced by the presence of hypoxic necrotizing lesions at treatment start. In contrast, RPT did not show a dose response when given at 5 and 10 mg/kg (Fig. 3b). The bacterial load of lungs in groups receiving 20 mg/kg RIF, 5 mg/kg RPT and 10 mg/kg RPT were roughly equivalent in absence of necrosis and hypoxia (Fig. 3b, left panel). RIF was more active in patient-like lesions in Nos2 −/− mice while RPT efficacy did not change with pathology: the burden of pulmonary TB at the end of the experiment was comparable in mice treated with 10 mg/kg RIF and 10 mg/kg RPT (Fig. 3b, right panel) principally resembling clinical outcome8. Organ homogenates plated on agar containing either RIF or RPT did not result in any c.f.u., confirming absence of genetic drug resistance. Taken together, RIF and RPT had similar efficacy in Nos2 −/− mice with hypoxic necrotizing lung lesions.

Pretomanid and delamanid but not metronidazole are bactericidal in hypoxic necrotizing lesions

The antibiotic metronidazole (MTZ) can kill non-replicating M. tuberculosis under hypoxic in vitro conditions, though only at very high concentrations that are hardly achieved in vivo 34. Daily administration of 400 mg/kg MTZ to Nos2 −/− mice had no effect on c.f.u. recovered from lung homogenates (Fig. 3c). A minute reduction after onset of necrosis and hypoxia in lung lesions (Fig. 3c) may be taken as indication for a hypoxia induced non-replicating M. tuberculosis subpopulation. In contrast, two chemically related derivatives, pretomanid (PTM) and delamanid (DLM), were bactericidal in Nos2 −/− mice with or without hypoxic necrotizing lesions (Fig. 3d,e). PTM is in advanced clinical development and DLM has recently been approved for use in patients with multidrug-resistant TB. Both drugs are bactericidal for replicating and non-replicating M. tuberculosis in culture and in mice that do not develop hypoxic necrotizing TB lesions35, 36. Our data support further development of PTN and DLM for TB chemotherapy.

Hypoxic necrotizing lesions render linezolid bacteriostatic, reduce activity of moxifloxacin and do not affect efficacy of BTZ043

The antimicrobial linezolid (LZD) is in clinical use for treatment of serious infections caused by Gram-positive pathogens including vancomycin-resistant enterococci, methicillin-resistant Staphylococcus aureus and penicillin-resistant Streptococcus pneumoniae 37. The drug is bactericidal in the BALB/c mouse model of TB, in C3HeB/FeJ mice with patient-like pathology and in patients with extensively drug-resistant TB13, 38. At a daily dose of 100 mg/kg given to Nos2 −/− mice for 4 weeks, LZD caused only marginal statistically insignificant reduction of lung c.f.u., as compared to treatment start (Fig. 3f). The presence of hypoxic necrotizing lesions abolished this trend suggesting bacteriostatic activity of LZD in this model (Fig. 3f).

The fourth generation fluoroquinolone, moxifloxacin (MXF), is broadly used for treatment of bronchitis, pneumonia and bacterial infections of the sinus, skin and abdomen. MXF has a high potency against M. tuberculosis in cultures and in mice without human-like lung pathology39. In our experiments, onset of hypoxia and necrosis in lung lesions of Nos2 −/− mice caused a 10-fold reduction of MXF bactericidal activity (Fig. 3g), indicating a direct impact of human-like pathology on MXF activity. The benzothiazinone pre-clinical drug candidate BTZ043, targeting M. tuberculosis cell wall biosynthesis, has nanomolar whole-cell activity and good in vivo efficacy in mice40. Unlike MFX, BTZ043 was highly potent in Nos2 −/− mice regardless of patient-like lung pathology (Fig. 3h).

Chemotherapy alters TB lung pathology in Nos2 −/−

For most drugs the impact of monotherapy on bacterial survival correlated fairly well with histopathological findings (Fig. 4). With the exception of MTZ, none of the study drugs facilitated development of advanced pathology when chemotherapy was initiated prior onset of hypoxia and necrosis (Fig. 4). Unlike RIF and BTZ043, the capacity of INH, RBT, PTM, DLM, LZD and MXF to resolve lung lesions was reduced (Fig. 4). MTZ closely resembled lesion occurrence in untreated animals and thus completely failed to alter pathology (Fig. 4c; Fig. 1c-d). INH, PTM, DLM and MXF maintained a small number of hypoxic necrotizing lesions until end of the experiment at day 84 eventually providing a niche for non-replicating pathogen survival (Fig. 4a,c-d,f). Only RIF and BTZ043 completely abolished hypoxia and necrosis in line with the strong bactericidal activity of both drugs (Fig. 3b,h; Fig. 4b,h). Intriguingly, LZD, bacteriostatic (Fig. 3f) in the presence of human-like lesions, significantly reduced occurrence of hypoxia, necrosis and overall numbers of lesions (Fig. 4f, Supplementary Fig. S2), which leaves the possibility that LZD may unfold more pronounced bactericidal activity after resolving advanced pathology.

Figure 4
figure 4

Impact of monotherapy on pulmonary pathology in Nos2 −/− mice with TB. Monotherapy for 4 weeks was initiated at day 42 (control group) or after onset of central necrosis and hypoxia in lung lesions at day 56 (study group) as described in Fig. 1a. Drugs were administered by oral gavage on 6 days per week. Consecutive thin sections of the formalin-fixed paraffin-embedded entire left lung lobe were stained with H&E or the pimonidazole hypoxia detection method and analyzed microscopically. One representative dataset out of 2 independent experiments is shown. Data were analyzed using two-way ANOVA with multicomparison and Tukey’s post-test. Shown are means ± s.e.m. and statistical significance as compared to the control group prior to drug treatment, n = 5, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. (a) isoniazid, 25 mg/kg. (b) rifampin, 10 mg/kg and 20 mg/kg; rifapentine, 5 mg/kg and 10 mg/kg. Statistical significance for different doses of the same drug at respective time points was similar and was only plotted once. (c) metronidazole, 400 mg/kg. (d) pretomanid, 75 mg/kg. (e) delamanid, 1 mg/kg. (f) linezolid, 100 mg/kg. (g) moxifloxacin, 200 mg/kg. (h) BTZ043, 50 mg/kg.

Full size image

Discussion

The Nos2 −/− mouse model of TB evaluated here displays significantly improved lung pathology over conventional BALB/c mice, comprising well-structured lesions with central necrosis and hypoxia as well as peripheral fibrosis (Supplementary Fig S1, Fig. 2). Following dermal infection, inflammatory foci in lungs progressed towards more advanced pathology in all infected mice, showing first signs of central necrosis followed by hypoxia, advanced necrosis and finally caseation in distinct lesions (Fig. 1c-d, Fig. 2). Pimonidazole adducts as indicators of hypoxia frequently appeared annular in the lesional centre indicating advanced necrosis and absence of viable cells capable of metabolizing the hydroxyprobe in the innermost centre. Our model presents similarities with the C3HeB/FeJ mouse model, but has the advantage of consistent and reproducible progression towards necrotic hypoxic lesions in all infected mice. M. tuberculosis infection of C3HeB/FeJ mice induces three distinct types of lesions: large patient-like lesions (Type I), rapidly progressive granulocytic lesions mainly composed of neutrophils (Type II) and small cellular inflammatory lesions (Type III, ‘BALB/c-type’)18. The heterogeneous distribution of lesion types and their differential responses to chemotherapy, as well as the early mortality of mice predominantly suffering from Type II lesions, remain major challenges of the C3HeB/FeJ mouse model of TB14. Type I lesions in C3HeB/FeJ mice were reported to develop liquefaction and even cavitation of necrotic lesions13, two pathological end-stage correlates of human TB that were not observed in Nos2 −/− mice before termination of our experiments at 84 days after infection (Supplementary Fig. S1). At this late stage of infection, a layer of infected foamy macrophages located on the inner side of the fibrotic capsule constitutes the interface between the cellular and caseous regions as seen in human granulomas26 (Fig. 2). Foamy macrophages harbour large quantities of lipids including cholesterol41 and presumably undergo necrosis thereby contributing lipids to the caseous centre. Cholesterol is likely one of the major carbon sources of mycobacteria during infection as suggested by transcriptome data42.

The primary aim of the current study was to evaluate the suitability of Nos2 −/− mice for TB drug development. Analysis of pathology over time revealed that lung lesions were homogeneously distributed and consistently developed central necrosis and hypoxia. Because the time window for onset of patient-like pathology is well defined between 42 and 56 days post-infection, we initiated drug administration at 42 days in the control group and at 56 days in the study group to specifically determine the impact of hypoxic necrotizing lesions on the outcome of monotherapy (Fig. 1a). A panel of nine drugs, either in clinical use or under clinical development was chosen for which efficacy data in humans or animal models are already available. The first-line TB drug INH showed low bactericidal activity in hypoxic necrotizing lesions (Fig. 3a). After an initial decline, the bacterial burden in lungs remained constant indicating the presence of drug-tolerant M. tuberculosis as validated by absence of genetic drug resistance after plating organ homogenates on agar supplemented with INH. It is tempting to speculate that the drug-tolerant subpopulation might be in a non-replicating state during which susceptibility to the cell wall inhibitor INH is lost. Evidence is increasing that bacterial persistence is a state actively maintained by a pool of intracellular stress responses allowing for prompt adaptation43. Whether INH-induced persistence of Mtb in Nos2 −/− mice compares to classic non-replicating survival described in culture models27, 28 or indeed is actively maintained44 will be addressed in the context of our future work evaluating TB multi-drug regimens in the Nos2 −/− model.

The rifamycins RIF and RPT showed strongest bactericidal activity among all drugs tested in our experiments. RPT was more potent prior to the onset of necrosis and hypoxia, whereas both drugs demonstrated comparable efficacy once patient-like pathology had established itself (Fig. 3b). Other murine models have suggested that RPT has roughly a 4-fold greater potency than RIF15. Our experiments indicate that hypoxic necrotizing lesions of Nos2 −/− mice render RIF more potent but do not affect RPT efficacy (Fig. 3b). Intriguingly, RIF has been found to accumulate in the caseum of human TB lesions over time21. This could explain its increased efficacy in our experiments. In turn this would imply that RPT does not accumulate in lesions of Nos2 −/− mice or in patients, which remains to be verified experimentally. Thus, performance of RIF and RPT in the Nos2 −/− model closely mimics clinical outcome8.

Gradual oxygen depletion or nutrient starvation induces a state of non-replicating drug tolerance in M. tuberculosis 27, 28. Moreover, oxygen limitation renders M. tuberculosis susceptible to high concentrations of MTZ, which may not be achievable in humans at a sustainable level28. The drug was moderately active in late-stage TB lesions in Nos2 −/− mice (Fig. 3c) indicating that a hypoxic non-replicating subpopulation resides in advanced caseous lesions. MTZ is active against M. bovis in rabbits7 and nonhuman primates45. Except for the late chronic stage of TB in BALB/c mice this drug lacks demonstrable activity in murine models including C3HeB/FeJ mice13, 46. MTZ may have led to earlier sputum smear and culture conversion in a clinical trial but was too neurotoxic for long-term use47. Two newer nitroimidazoles, PTM and DLM exhibit mycobactericidal activity both in growing and in hypoxic non-replicating bacilli and hence are more promising candidates against M. tuberculosis 48. DLM has been approved for patients with multidrug resistant TB and PTM is currently undergoing late stage clinical development. Both drugs expressed comparable activity in hypoxic necrotizing lesions of Nos2 −/− mice (Fig. 3d,e) supporting their further evaluation.

The drug candidate, LZD, with demonstrable efficacy in extensively drug-resistant TB patients38, reduced the M. tuberculosis load in lungs of Nos2 −/− mice prior to onset of necrosis and hypoxia and was bacteriostatic thereafter (Fig. 3f). Surprisingly this drug reduced occurrence of lung lesions including necrosis and completely abolished hypoxia (Fig. 4f). LZD has been shown to have immunomodulatory and anti-inflammatory effects49, 50, which may have supported pathologic alterations observed in our experiments. In C3HeB/FeJ mice a linear reduction of the bacterial burden in lungs over 8 weeks of LZD monotherapy was observed originally13. Using larger experimental groups, however, it was found that C3HeB/FeJ mice segregate into two groups of LZD responders either showing reduced c.f.u. in lungs similar to BALB/c mice or revealing a bacteriostatic profile14. Our results are comparable to the latter group of C3HeB/FeJ mice likely characterized by large caseous necrotic lesions18.

In general, MFX was strongly bactericidal in Nos2 −/− mice, but roughly 10-fold less effective after onset of necrosis and hypoxia (Fig. 3g). Reduced activity was likely caused by insufficient lesion penetration of the drug as demonstrated in TB patients21. Several clinical trials to evaluate new TB drugs and treatment regimens have generated disappointing results8,9,10. The REMox TB trial, one of the largest clinical studies in the field, assessed whether replacement of INH or EMB in the standard regimen RIF/INH/EMB/PZA by the DNA-gyrase inhibitor MFX could shorten treatment time to 4 months. This trial was based on promising results from various murine models51,52,53 and on clinical data indicating a moderately higher rate of sputum conversion when EMB or INH were substituted by MFX54,55,56,57,58. Although initial murine studies failed to predict the clinical outcome of shorter MFX treatment time, carefully designed mouse experiments mimicked clinical results, underscoring the value of appropriately designed small rodent models for TB drug development59. TB drug testing in C3HeB/FeJ mice and new insights into drug penetration of TB lesions in patients and rabbits led to the conclusion that animal models showing patient-like TB pathology are key in overcoming the disconnect between pre-clinical and clinical studies60. In contrast to MFX, BTZ043 possessed consistent potency in Nos2 −/− mice regardless of pulmonary pathology (Fig. 4h) suggesting uniform tissue distribution and sufficient lesion penetration. The profound bactericidal activity of BTZ043 in the Nos2 −/− model, ranging between the most potent TB first-line drugs, INH and RIF, warrants further investigation of this and related benzothiazinones61.

In conclusion, unlike other murine models, Nos2 −/− mice with patient-like pathology in principle correctly predicted clinical outcomes of RIF/RBT8, MTZ47, MXF9, 10, and potentially LZD38 (Supplementary Table S1). Results obtained for PTM, DLM and BTZ043 in Nos2 −/− mice support development and consideration of these drugs for standard TB chemotherapy. Note that drugs interfering with host NO production or requiring activation by NO may not be suitable for testing in Nos2 −/− mice due to their inability to produce NO in immune cells, which represents a limitation of this model. The development of new drugs and regimens for shortening the treatment of TB requires understanding the impact of specific drugs on specific types of lesions coexisting in patients. TB lesions in humans are very complex and often form a continuum of multiple discrete types of pathologies ranging from pneumonic infiltrates to dense consolidations to cavitation. No single model will suffice for understanding the impact of a new drug or regimen on all these types of pathologies. Yet, the Nos2 −/− mouse with a patient-like pathology of lung granulomas and uniform disease progression provides a highly consistent and predictive model for this crucial type of pathology in TB drug development.

Methods

Bacterial strains and growth conditions

M. tuberculosis H37Rv (American Type Culture Collection, #27294) was grown in Middlebrook 7H9 broth (Becton Dickinson) supplemented with albumin-dextrose-catalase enrichment (Becton Dickinson), 0.2% glycerol, 0.05% Tween 80 or on Middlebrook 7H11 agar (Becton Dickinson) containing 10% v/v oleic acid-albumin-dextrose-catalase enrichment (Becton Dickinson) and 0.2% glycerol. Infection stocks were prepared from mid-log phase cultures. For c.f.u. determinations, serial dilutions were performed in PBS/0.05% Tween 80 and plated onto Middlebrook 7H11 agar. Plates were incubated at 37 °C for 3–4 weeks prior to c.f.u. counting.

Animal experiments

Female C57BL/6 Nos2−/− mice were bred in-house and maintained under specific pathogen-free conditions. Eight- to ten-week-old animals were anesthetized (ketamine 65 mg/kg, acepromazine 2 mg/kg, xylazine 11 mg/kg) and infected by injecting 1,000 c.f.u. of M. tuberculosis in 20 μl PBS into the ear dermis. At 14 and 21 days post-infection each mouse received 0.5 mg of monoclonal anti-tumour necrosis factor alpha antibody (purified from MP6-XT22 cultures) by i.p. injection. Two hours before euthanasia animals received 60 mg/kg pimonidazole hydrochloride i.p. to allow for detection of hypoxic regions in organ sections.

Ethical statement

All animal studies have been ethically reviewed and approved by the State Office for Health and Social Services, Berlin, Germany. Experimental procedures were carried out in accordance with the European directive 2010/63/EU on Care, Welfare and Treatment of Animals.

Drugs, formulations and administration

INH, LZD, RIF, RPT, MXF and MTZ were purchased from Sigma and formulated in 0.4% methylcellulose. Rifamycins were dissolved in dimethyl sulfoxide prior dilution in 0.4% methylcellulose. The final concentration of dimethyl sulfoxide did not exceed 5%. BTZ04340, PTM and DLM62, 63 were synthesized in-house and formulated in 0.25% carboxymethyl cellulose/0.05% Tween 80 (BTZ043) or in 10% hydroxypropyl-β-cyclodextrine/10% lecithin (PTM, DLM)35, 64. Drug formulations stored at 4 °C were administered in 0.2 ml per dose by oral gavage for 6 days per week.

Staining procedures and histopathology

The left lung lobe of mice was fixed using 4% paraformaldehyde in PBS and embedded in paraffin. Organ sections (2–3 μm) were deparaffinised and subjected to haematoxylin and eosin (H&E) staining, Trichrome staining (Sigma), hypoxia staining (Hypoxyprobe-1 kit, Hypoxyprobe Inc.) or Ellis staining (Sigma) to visualize acid-fast bacilli in tissues; staining was performed according to the manufacturer’s instructions. Tissue sections were analysed using a Zeiss Axio Imager Z1 with CCD AxioCam. A researcher blinded to the study groups scored 16 individual stained sections of each organ in study groups of five mice per time point. Central necrosis of lesions was defined as a lighter pink region indicating tissue consolidation surrounded by granulomatous inflammatory infiltrate. Lesions were considered hypoxic when pimonidazole adducts in their central region could be visualized by a brown stain.

Bacterial burden of mouse lungs and drug-resistant colonies

Mice were euthanized at dedicated time points and superior, middle inferior and post-caval lobes were removed and homogenized in 1 ml PBS/0.05% Tween 80. Serial dilutions of organ homogenates were plated onto Middlebrook 7H11 agar and in addition on agar supplemented with 0.4% activated charcoal for all time points during chemotherapy. Plates showing higher c.f.u. counts were used for data analysis. To determine the level of drug resistance at the last time point of each experiment, ¼ of the lung homogenate was plated onto Middlebrook 7H11 agar containing 4 μg/ml INH, 0.4 μg/ml RIF, 0.4 μg/ml RPT or 8 μg/ml LZD. Plates were examined after 4 weeks of incubation at 37 °C and kept until week 8 for re-examination. The limit of detection for drug-resistant colonies was 4 c.f.u. per organ.